From todd at phy.ucsf.edu  Mon Aug  1 12:30:59 1994
From: todd at phy.ucsf.edu (todd@phy.ucsf.edu)
Date: Mon, 1 Aug 1994 09:30:59 -0700
Subject: New paper: Ffwd Hebbian Learning w/ Nonlinear Outputs
Message-ID: <9408011630.AA04378@dizzy.ucsf.EDU>



FTP-host: archive.cis.ohio-state.edu
FTP-filename: /pub/neuroprose/troyer.ffwd_hebb.ps.Z

The following paper has been submitted to Neural Networks and is
available from the Ohio State neuroprose archive. 

TITLE:   Feedforward Hebbian Learning with Nonlinear Output Units:
         A Lyapunov Approach

AUTHOR:  Todd Troyer <todd at phy.ucsf.edu>

         W.M. Keck Center for Integrative Neuroscience
            and Department of Pysiology
         513 Parnassus Ave. Box 0444
         University of California, San Francisco
         San Francisco, CA.  94143

ABSTRACT:

A Lyapunov function is constructed for the unsupervised learning
equations of a large class of two layer networks.  Units in the output
layer are recurrently connected by fixed symmetric weights; only the
feedforward connections between layers undergo learning.  In contrast
to much of the previous work on self-organization of this type, the
output units have nonlinear transfer functions.  The Lyapunov function
is similar in form to that derived by Cohen-Grossberg and Hopfield.
Two theorems are proved regarding the location of stable equilibria in
the limit of high gain transfer functions.  The analysis is applied to
the soft competitive learning networks of Amari and Takeuchi.


Retrieve this paper by anonymous ftp from:
archive.cis.ohio-state.edu (128.146.8.52)
    in the /pub/neuroprose    directory

The name of the paper in this archive is:
    troyer.ffwd_hebb.ps.Z          [15 pages]

No hard copies available.




From giles at research.nj.nec.com  Mon Aug  1 13:12:11 1994
From: giles at research.nj.nec.com (Lee Giles)
Date: Mon, 1 Aug 94 13:12:11 EDT
Subject: Reprint:SYNAPTIC NOISE IN DYNAMICALLY-DRIVEN RECURRENT NEURAL NETWORKS
Message-ID: <9408011712.AA13061@fuzzy>



****************************************************************************************

Reprint:SYNAPTIC NOISE IN DYNAMICALLY-DRIVEN RECURRENT NEURAL NETWORKS:
        CONVERGENCE AND GENERALIZATION

The following reprint is available via the University of Maryland Department of Computer 
Science Technical Report archive:

_________________________________________________________________________________________

          "Synaptic Noise in Dynamically-driven Recurrent Neural Networks: 
                       Convergence and Generalization"

      UNIVERSITY OF MARYLAND TECHNICAL REPORT UMIACS-TR-94-89 AND CS-TR-3322
               
                     Kam Jim(a), C.L. Giles(a,b), B.G. Horne(a)

                     {kamjim,giles,horne}@research.nj.nec.com

        (a) NEC Research Institute,4 Independence Way, Princeton, NJ 08540
  (b) Institute for Advanced Computer Studies, U. of Maryland, College Park, MD 20742  


There has been much interest in applying noise to feedforward neural networks in order 
to observe their effect on network performance. We extend these results by introducing 
and analyzing various methods of injecting synaptic noise into dynamically-driven 
recurrent networks during training. By analyzing and comparing the effects of these 
noise models on the error function, we found that applying a controlled amount of noise 
during training can improve convergence time and generalization performance. In addition, 
we analyze the effects of various noise parameters (additive vs. multiplicative, 
cumulative vs. non-cumulative, per time step vs. per sequence) and predict that best 
overall performance can be achieved by injecting additive noise at each time step. Noise 
contributes a second-order gradient term to the error function which can be viewed as an 
anticipatory agent to aid convergence. This term appears to find promising regions of 
weight space in the beginning stages of training when the training error is large and 
should improve convergence on error surfaces with local minima.Synaptic noise also 
enhances the error function by favoring internal representations where state nodes are 
operating in the saturated regions of the sigmoid discriminant function, thus improving 
generalization to longer sequences. We substantiate these predictions by performing 
extensive simulations on learning the dual parity grammar from grammatical strings 
encoded as temporal sequences with a second-order fully recurrent neural network.

--------------------------------------------------------------------------------------

                          FTP INSTRUCTIONS

                unix> ftp cs.umd.edu (128.8.128.8)
                Name: anonymous
                Password: (your_userid at your_site)
                ftp> cd pub/papers/TRs
                ftp> binary
                ftp> get 3322.ps.Z
                ftp> quit
                unix> uncompress 3322.ps.Z

--------------------------------------------------------------------------------------


--                                 
C. Lee Giles / NEC Research Institute / 4 Independence Way
Princeton, NJ 08540 / 609-951-2642 / Fax 2482
==




From inbs at bingsuns.cc.binghamton.edu  Mon Aug  1 13:48:58 1994
From: inbs at bingsuns.cc.binghamton.edu (INBS-conference)
Date: Mon, 1 Aug 94 13:48:58 EDT
Subject: call for papers
Message-ID: <9408011748.AA23972@bingsuns.cc.binghamton.edu >


                 C A L L        for      P A P E R S

          First International IEEE Symposium on  


       INTELLIGENCE in NEURAL AND BIOLOGICAL SYSTEMS


          May 29-31,1995, Washington DC Area, USA
        
           Sponsored by: IEEE Computer Society; 


In Cooperating with: AAAS Society; AAAI Society; SMC Society


AIMS and SCOPE of the Symposium

The Intelligence in Artificial Neural Networks and the Computational 
evolution of the Biological Systems are two very important and very 
active research areas, which offer and promise many practical 
applications to scientists and other professionals in industry 
and goverment as well. In response to this demand, the INBS 
Symposium offers a theoretical and a practical medium for the 
evolutionary and the intelligence processes in both artificial and 
biological systems and the interaction between these fields.

Some Topics 

 EVOLUTIONARY COMPUTING (DNA sequence processing, Genome Processes,  
  DNA, Topologies, Synthesis of DNA, Formal Linguistic of DNA, 
  Structure/Function Correlation,   Computational Genetics)
 BIOTECHNOLOGY AND APPLICATIONS (Mapping the Genome, Human Genome,      
  Molecular Computing, Limitations of Biological Models)
 GENETIC ALGORITHMS (Clustering, Optimization, Searching, Programming, etc)
 LANGUAGES UNDERSTANDING (Natural Languages, NL Translation, 
  Text Abstraction, Computational Linguistics)
 LEARNING AND PERCEPTION (Supervised, Unsupervised, Hybrid, 
  Understanding, Planning, Interpretation)
 NEUROSCIENCE (Adaptive Control Models, Fuzzy & Probabilistic Models, 
  Hybrid Models, Dynamic Neural and Neurocomputing Models, 
  Self-Organized Models)

SOME KEYNOTE DISTIGUISHED SPEAKERS

A.Apostolico, Europe;  S.Arikawa, Japan; S.Hameroff, USA


PROGRAM COMMITTEE                      Local Arrangement/Publicity Committee
N.Bourbakis, BU-SUNY,USA, Chair        Cynthia Shapiro
R.Brause, UG,Germany                   Sukarno Mertoguno
K.DeJong, GMU,USA                      James Gattiker
J.Shavlik, UWM,USA                     Ali Moghaddamzadeh
C.Koutsougeras, TU,USA
H.Kitano, JAPAN                                
M.Perlin, CMU,USA                      Publication Registration Committee
H.Pattee, BU,USA                       D.I. Kavraki
D.Schaffer, Philips Lab,USA                     
D.Searls, UPA,USA                               
J.Collado-Vides, UNAM
T.Yokomori, UEC,Japan
S.Rasmussen, Los Alamos,NL
G.Paun, Roumania
A.Restivo, U.Palermo,Italy
M.Chrochmore, U.Paris, France
D.Perrin, U.Paris, France
R.Reynolds, WSU,USA
M.Conrad, WSU,USA
M.Kameyama, TU,Japan
J.Nikolis, UP,Greece
T.Head, BU, USA, Vice Chair
C.Benham, MS,USA
R.VanBuskirk, BU-SUNY,USA
E.Dietrich, BU-SUNY,USA
S.Koslow, NIH,USA
M.Huerta, NIH,USA
B.Punch, MSU,USA



INFORMATION FOR AUTHORS
Authors are requested to submit four copies (in English) of 
their typed complete manuscript (25 pages max), or an 
extended summary (5-10 pages max) by Nov. 31,1994, to 
N. G. Bourbakis, Binghamton University,T.J.Watson School,
AAAI Lab, NY 13902,Tel: 607-777-2165, 607-771-4033; 
Fax:607-777-4822, E-mail : Bourbaki at BingSuns.CC.Binghamton.edu, 
or INBS at Bingsuns.CC.Binghamton.edu.
Each manuscript submitted to INBS must indicate the most relevant 
areas and include the complete address of at least one of the 
authors. Notification of acceptance, Jan. 31,1995. Final
copies of the papers accepted by INBS due March 21 ,1995.


From tesauro at watson.ibm.com  Mon Aug  1 13:50:48 1994
From: tesauro at watson.ibm.com (tesauro@watson.ibm.com)
Date: Mon, 1 Aug 94 13:50:48 EDT
Subject: NIPS*94 Registration
Message-ID: <mailman.694.1149540293.24850.connectionists@cs.cmu.edu>

Registration for NIPS*94 is now open.  A registration brochure
is available on-line via the NIPS*94 Mosaic homepage
(http://www.cs.cmu.edu:8001/afs/cs/project/cnbc/nips/NIPS.html),
or by anonymous FTP:
FTP site: mines.colorado.edu (138.67.1.3)
FTP file: /pub/nips94/nips94-registration-brochure.ps

The brochure contains the registration form and describes
the program highlights, including the list of invited speakers
and tutorial speakers.

People without access to FTP or Mosaic may request a copy of
the brochure by e-mail to "nips94 at mines.colorado.edu" or by
physical mail to:

        NIPS*94 Registration
        Dept. of Mathematical and Computer Sciences
        Colorado School of Mines
        Golden, CO  80401  USA


-- Gerry Tesauro
   NIPS*94 General Chair

From plaut at cmu.edu  Tue Aug  2 10:14:54 1994
From: plaut at cmu.edu (David Plaut)
Date: Tue, 02 Aug 1994 10:14:54 -0400
Subject: TR: Understanding Normal and Impaired Word Reading
Message-ID: <7478.775836894@crab.psy.cmu.edu>

FTP-host:	hydra.psy.cmu.edu [128.2.248.152]
FTP-file:	pub/pdp.cns/pdp.cns.94.5.ps.Z

[78 pages; 353Kb compressed; 924Kb uncompressed]

For those who do not have FTP access, physical copies can be requested from
Barbara Dorney <bd1q+ at andrew.cmu.edu>.

============================================================================

		Understanding Normal and Impaired Word Reading:
	       Computational Principles in Quasi-Regular Domains

	     David C. Plaut                    James L. McClelland
	Carnegie Mellon University          Carnegie Mellon University

	    Mark S. Seidenberg                 Karalyn E. Patterson
      University of Southern California     MRC Applied Psychology Unit

			 Technical Report PDP.CNS.94.5
				   July 1994

We develop a connectionist approach to processing in quasi-regular domains, as
exemplified by English word reading.  A consideration of the shortcomings of a
previous implementation (Seidenberg & McClelland, 1989, Psych. Rev.)  in
reading nonwords leads to the development of orthographic and phonological
representations that capture better the relevant structure among the written
and spoken forms of words.  In a number of simulation experiments, networks
using the new representations learn to read both regular and exception words,
including low-frequency exception words, and yet are still able to read
pronounceable nonwords as well as skilled readers.  A mathematical analysis of
the effects of word frequency and spelling-sound consistency in a related but
simpler system serves to clarify the close relationship of these factors in
influencing naming latencies.  These insights are verified in subsequent
simulations, including an attractor network that reproduces the naming latency
data directly in its time to settle on a response.  Further analyses of the
network's ability to reproduce data on impaired reading in surface dyslexia
support a view of the reading system that incorporates a graded
division-of-labor between semantic and phonological processes.  Such a view is
consistent with the more general Seidenberg and McClelland framework and has
some similarities with---but also important differences from---the standard
dual-route account.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
David Plaut			plaut at cmu.edu	      "Doubt is not a pleasant
Department of Psychology	412/268-5145	       condition, but certainty
Carnegie Mellon University	412/268-5060 (FAX)     is an absurd one."
Pittsburgh, PA  15213-3890	345H Baker Hall			--Voltaire
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

From john at dcs.rhbnc.ac.uk  Tue Aug  2 11:12:04 1994
From: john at dcs.rhbnc.ac.uk (john@dcs.rhbnc.ac.uk)
Date: Tue, 02 Aug 94 16:12:04 +0100
Subject: Technical Report Series in Neural and Computational Learning
Message-ID: <8129.9408021512@platon.cs.rhbnc.ac.uk>


The European Community ESPRIT Working Group in Neural and Computational 
       Learning Theory (NeuroCOLT):

---------------------------------------
NeuroCOLT Technical Report NC-TR-94-5:
---------------------------------------
A Weak Version of the Blum, Shub \& Smale Model

by Pascal Koiran

Abstract:
We propose a weak version of the Blum-Shub-Smale model of computation 
over the real numbers. In this weak model only a ``moderate" usage 
of multiplications and divisions is allowed. 
The class of boolean languages recognizable in polynomial
time is shown to be the complexity class P/poly. The main tool is a result
on the existence of small rational points in semi-algebraic sets which is
of independent interest. As an application, we generalize recent results 
of Siegelmann \& Sontag on recurrent neural networks, and of Maass 
on feedforward nets. A preliminary version of this paper was presented at
the 1993 IEEE Symposium on Foundations of Computer Science. Additional
results include:
\begin{itemize}
\item an efficient simulation of order-free real Turing machines by
probabilistic Turing machines in the full Blum-Shub-Smale model;
\item an efficient simulation of arithmetic circuits over the integers by
boolean circuits;
\item the strict inclusion of the real polynomial hierarchy in
weak exponential time.
\end{itemize}


------------------------
The Report NC-TR-94-5 can be accessed and printed as follows 

% ftp cscx.cs.rhbnc.ac.uk  (134.219.200.45)
Name: anonymous
password: your full email address
ftp> cd pub/neurocolt/tech_reports
ftp> binary
ftp> get nc-tr-94-5.ps.Z
ftp> bye
% zcat nc-tr-94-5.ps.Z | lpr -l

Uncompressed versions of the postscript files have also been left for anyone not
having an uncompress facility.

Best wishes
John Shawe-Taylor





From sutton at gte.com  Tue Aug  2 15:54:04 1994
From: sutton at gte.com (Rich Sutton)
Date: Tue, 2 Aug 1994 14:54:04 -0500
Subject: On Step-Size and Bias in TD Learning (Paper Available)
Message-ID: <199408021849.AA05405@ns.gte.com>

The following paper appeared in the proceeding of the 1994 Yale Workshop on
Adaptive and Learning Systems, 1994.  ftp instructions follow.


        ON STEP-SIZE AND BIAS IN TEMPORAL-DIFFERENCE LEARNING

           by Richard S. Sutton and Satinder P. Singh (MIT)
                                       
We present results for three new algorithms for setting the step-size
parameters, alpha and lambda, of temporal-difference learning methods such
as TD(lambda).  The overall task is that of learning to predict the outcome
of an unknown Markov chain based on repeated observations of its state
trajectories. The new algorithms select step-size parameters online in such
a way as to eliminate the bias normally inherent in temporal-difference
methods. We compare our algorithms with conventional Monte Carlo methods.
Monte Carlo methods have a natural way of setting the step size: for each
state s they use a step size of  1/n(s), where n(s) is the number of times
state s has been visited.  We seek and come close to achieving comparable
step-size algorithms for TD(lambda). One new algorithm uses a lambda=1/n(s)
schedule to achieve the same effect as processing a state backwards with
TD(0), but remains completely incremental. Another algorithm uses a lambda
at each time equal to the estimated transition probability of the current
transition. We present empirical results showing improvement in convergence
rate over Monte Carlo methods and conventional TD(lambda).  A limitation of
our results at present is that they apply only to tasks whose state
trajectories do not contain cycles.  

================================================================
unix> ftp envy.cs.umass.edu
 
Name: anonymous
Password: [your ID]
ftp> cd pub/singh
ftp> binary
ftp> get Yale94.ps.Z
ftp> bye
 
unix> uncompress Yale94.ps.Z
unix> [your command to print PostScript file] Yale94.ps
===============================================================
The paper is six pages long.  The file is about 130K.

FTP-host: envy.cs.umass.edu
FTP-filename: /pub/singh/Yale94.ps.Z



From marwan at sedal.su.oz.au  Tue Aug  2 23:12:08 1994
From: marwan at sedal.su.oz.au (Marwan Jabri)
Date: Wed, 3 Aug 1994 13:12:08 +1000
Subject: paper available: Practical Performance and Credit Assignment...
Message-ID: <199408030312.NAA28697@sedal.sedal.su.OZ.AU>

FTP-host: archive.cis.ohio-state.edu
FTP-filename: /pub/neuroprose/jabri.ppap.ps.Z

The file jabri.ppap.ps.Z is now available for
copying from the Neuroprose repository and has been 
submitted for publication:


	Practical Performance and Credit Assignment Efficiency 
	of Analog Multi-layer Perceptron Perturbation Based 
	Training Algorithms

	Marwan A. Jabri

	Systems Engineering and Design Automation Laboratory 
	Sydney University Electrical Engineering
	NSW 2006 Australia
	marwan at sedal.su.oz.au

	SEDAL Technical Report 1-7-94


Abstract
Many algorithms have been recently reported for the training of analog
multi-layer perceptron. Most of these algorithms were evaluated either
from a computational or simulation view point. This paper applies several
of these algorithms to the training of an analog multi-layer perceptron
chip. The advantages and shortcomings of these algorithms in terms of
training and generalisation performance and their capabilities in a
limited precision environment are discussed. Extensive experiments
demonstrate that a trade-off exists between the parallelisation of
perturbations and the efficiency of credit assignment. Two
semi-parallelisation heuristics are presented and are shown to provide
advantages in terms of efficient exploration of the solution space and
fewer credit assignment confusions.


Retrieve this paper by anonymous ftp from:
archive.cis.ohio-state.edu (128.146.8.52)
    in the /pub/neuroprose    directory

The name of the paper in this archive is:
    jabri.ppap.ps.Z          [28 pages]


Sorry, but no hard copies available.

From niranjan at eng.cam.ac.uk  Wed Aug  3 15:07:20 1994
From: niranjan at eng.cam.ac.uk (niranjan@eng.cam.ac.uk)
Date: Wed, 3 Aug 94 15:07:20 BST
Subject: Post Doc job in Cambridge
Message-ID: <18000.9408031407@tulip.eng.cam.ac.uk>

---------------------- JOB  JOB  JOB  JOB -------------------------

    University Of Cambridge, Department of Obstetrics & Gynaecology

	    Post-Doctoral Research Assistant

    Neural Networks in the Quality Assurance in Maternity Care  (QAMC)

Under the terms of a grant recently awarded to the QAMC project by the
Commission of the European Communities (CEC), we expect that we will
soon be able to offer a Post Doctoral Research Position in Cambridge
for the above investigation.


From g.gaskell at psychology.bbk.ac.uk  Thu Aug  4 14:17:00 1994
From: g.gaskell at psychology.bbk.ac.uk (Gareth)
Date: Thu, 4 Aug 94 14:17 BST
Subject: Thesis: Speech Perception & Connectionsim
Message-ID: <mailman.697.1149540293.24850.connectionists@cs.cmu.edu>

(Sorry - ignore the previous message, I sent the wrong
file!)

FTP-host: archive.cis.ohio-state.edu (128.146.8.52)
FTP-filename: /pub/neuroprose/Thesis/gaskell.thesis.ps.Z

A new PhD thesis (150 pages) is now available in the 
neuroprose archive.  The thesis examines the role of 
phonological variation in human speech perception using both 
experimental and connectionist techniques.

Abstract:

The research reported in this thesis examines issues of word 
recognition in human speech perception.  The main aim of the 
research is to assess the effect of regular variation in speech on 
lexical access.  In particular, the effect of a type of neutralising 
phonological variation, assimilation of place of articulation, is 
examined.  This variation occurs regressively across word 
boundaries in connected speech, altering the surface phonetic 
form of the underlying words.  Two methods of investigation 
are used to explore this issue.  Firstly, experiments using cross-
modal priming and phoneme monitoring techniques are used to 
examine the effect of variation on the matching process between 
speech input and lexical form.  Secondly, simulated experiments 
are performed using two computational models of speech 
recognition: TRACE (McClelland & Elman, 1986) and a simple 
recurrent network.

The priming experiments show that the mismatching effects of a 
phonological change on the word-recognition process depend 
on their viability, as defined by phonological constraints.  This 
implies that speech perception involves a process of context-
dependent inference, that recovers the abstract underlying 
representation of speech.  Simulations of these and other 
experiments are then reported using a simple recurrent network 
model of speech perception.  The model accommodates the 
results of the priming studies and predicts that similar 
phonological context effects will occur in non-words.  Two 
phoneme monitoring studies support this prediction, but also 
show interaction between lexical status and viability, implying 
that phonological inference relies on both lexical and 
phonological constraints.  A revision of the network model is 
proposed which learns the mapping from the surface form of 
speech to semantic and phonological representations.

To retrieve the file:

ftp archive.cis.ohio-state.edu
login: anonymous
password:<email address>
ftp> cd /pub/neuroprose/Thesis
ftp> binary
ftp> get gaskell.thesis.ps.Z
ftp> bye
uncompress gaskell.thesis.ps.Z
lpr gaskell.thesis.ps [or whatever you normally do to print]

Gareth Gaskell
Centre for Speech and Language,
Birkbeck College,
London,

g.gaskell at psyc.bbk.ac.uk


From giles at research.nj.nec.com  Thu Aug  4 18:00:26 1994
From: giles at research.nj.nec.com (Lee Giles)
Date: Thu, 4 Aug 94 18:00:26 EDT
Subject: Reprint:LEARNING A CLASS OF LARGE FINITE STATE MACHINES WITH A RECURRENT 
Message-ID: <9408042200.AA18471@fuzzy>

NEURAL NETWORK 

The following reprint is available via the University of Maryland Department 
of Computer Science and the NEC Research Institute archives:

_____________________________________________________________________________



                 Learning a Class of Large Finite State Machines
                      with a Recurrent Neural Network

        UNIVERSITY OF MARYLAND TECHNICAL REPORT UMIACS-TR-94-94 AND CS-TR-3328

                 C. L. Giles[1,2], B. G. Horne[1], T. Lin[1,3] 

      [1] NEC Research Institute, 4 Independence Way, Princeton, NJ  08540
            [2]UMIACS, University of Maryland, College Park, MD 20742
          [3] EE Department, Princeton University, Princeton, NJ 08540

                      {giles,horne,lin}@research.nj.nec.com
                
One of the issues in any learning model is how it scales with problem size.
Neural networks have not been immune to scaling issues. We show that a
dynamically- driven discrete-time recurrent network (DRNN) can learn rather
large grammatical inference problems when the strings of a finite memory
machine (FMM) are encoded as temporal sequences. FMMs are a subclass of
finite state machines which have a finite memory or a finite order of
inputs and outputs. The DRNN that learns the FMM is a neural network that
maps directly from the sequential machine implementation of the FMM. It has
feedback only from the output and not from any hidden units; an example is
the recurrent network of Narendra and Parthasarathy. (FMMs that have zero
order in the feedback of outputs are called definite memory machines and
are analogous to Time-delay or Finite Impulse Response neural networks.)
Due to their topology these DRNNs are as least as powerful as any
sequential machine implementation of a FMM and should be capable of
representing any FMM. We choose to learn ``particular FMMs.\' Specif
ically, these FMMs have a large number of states (simulations are for $256$
and $512$ state FMMs) but have minimal order, relatively small depth and
little logic when the FMM is implemented as a sequential machine.
Simulations for the num ber of training examples versus generalization
performance and FMM extraction size show that the number of training
samples necessary for perfect generalization is less than that necessary to
completely characterize the FMM to be learned. This is in a sense a best
case learning problem since any arbitrarily chosen FMM with a minimal
number of states would have much more order and string depth and most
likely require more logic in its sequential machine implementation.


--------------------------------------------------------------------------

                          FTP INSTRUCTIONS

                unix> ftp cs.umd.edu (128.8.128.8)
                Name: anonymous
                Password: (your_userid at your_site)
                ftp> cd pub/papers/TRs
                ftp> binary
                ftp> get 3328.ps.Z
                ftp> quit
                unix> uncompress 3328.ps.Z

                                 OR

                unix> ftp external.nj.nec.com (138.15.10.100)
                Name: anonymous
                Password: (your_userid at your_site)
                ftp> cd pub/giles/papers
                ftp> binary
                ftp> get large.fsm.ps.Z
                ftp> quit
                unix> uncompress large.fsm.ps.Z

--------------------------------------------------------------------------

--                                 
C. Lee Giles / NEC Research Institute / 4 Independence Way
Princeton, NJ 08540 / 609-951-2642 / Fax 2482
==




From john at dcs.rhbnc.ac.uk  Fri Aug  5 08:00:06 1994
From: john at dcs.rhbnc.ac.uk (john@dcs.rhbnc.ac.uk)
Date: Fri, 05 Aug 94 13:00:06 +0100
Subject: Technical Report Series in Neural and Computational Learning
Message-ID: <10057.9408051200@platon.cs.rhbnc.ac.uk>


The European Community ESPRIT Working Group in Neural and Computational 
       Learning Theory (NeuroCOLT):

Many apologies that the previous technical report NC-TR-94-5 was not available
when first announced. It is now installed. I have also installed an ascii file
entitled `abstracts' which holds an up to date list of the Technical Reports
with their abstracts.

A new technical report is also now available:
---------------------------------------
NeuroCOLT Technical Report NC-TR-94-7:
---------------------------------------
On the power of real Turing machines over binary inputs
by Felipe Cucker, Universitat Pompeu Fabra, Balmes 132, Barcelona 08008, SPAIN
and Dima Grigoriev, Depts. of Comp. Science and Maths, Penn State University,
    University Park, PA 16802, USA

Abstract:

In recent years the study of the complexity of computational problems
involving real numbers has been an increasing research area.  Blum,
Shub and Smale (1989) proposed a computational model ---the real Turing
machine--- for dealing with the above problems was developed.

The aim of this paper is to prove that
$\BP(\PAR)=\;$PSPACE/{\it poly} where $\PAR$ is the class of sets
computed in parallel polynomial time by (ordinary) real Turing machines.
As a consequence we obtain the existence of binary sets that do not
belong to the
Boolean part of $\PAR$ (an extension of the result of Koiran (1994)
since $\PH\subseteq\PAR$) and a separation of complexity classes in
the real setting.

------------------------
The Report NC-TR-94-7 can be accessed and printed as follows (follow the same
procedure for getting the abstracts file)

% ftp cscx.cs.rhbnc.ac.uk  (134.219.200.45)
Name: anonymous
password: your full email address
ftp> cd pub/neurocolt/tech_reports
ftp> binary
ftp> get nc-tr-94-7.ps.Z
ftp> bye
% zcat nc-tr-94-7.ps.Z | lpr -l

Uncompressed versions of the postscript files have also been left for anyone not
having an uncompress facility.

Best wishes
John





From shultz at hebb.psych.mcgill.ca  Fri Aug  5 14:19:36 1994
From: shultz at hebb.psych.mcgill.ca (Tom Shultz)
Date: Fri, 5 Aug 94 14:19:36 EDT
Subject: No subject
Message-ID: <9408051819.AA06064@hebb.psych.mcgill.ca>

Subject: Paper available: Modeling cognitive development on balance 
scale phenomena
Date: 5 Aug. '94

FTP-host: archive.cis.ohio-state.edu
FTP-file: pub/neuroprose/shultz.balance-ml.ps.Z

-------------------------------------------------------------

The following paper has been placed in the Neuroprose archive at 
Ohio State University: 

Modeling cognitive development on balance scale phenomena (33 
pages)

Thomas R. Shultz, Denis Mareschal, & William C. Schmidt
Department of Psychology & McGill Cognitive Science Centre		
McGill University
Montreal, Quebec, Canada H3A 1B1		
shultz at psych.mcgill.ca

Abstract
We used cascade-correlation to model human cognitive development 
on a well studied psychological task, the balance scale. In balance 
scale experiments, the child is asked to predict the outcome of 
placing certain numbers of equal weights at various distances to the 
left or right of a fulcrum. Both stage progressions and information 
salience effects have been found with children on this task. 
Cascade-correlation is a generative connectionist algorithm that 
constructs its own network topology as it learns. Cascade-
correlation networks provided better fits to these human data than 
did previous models, whether rule-based or connectionist. The 
network model was used to generate a variety of novel predictions 
for psychological research.

Keywords: cognitive development, balance scale, connectionist 
learning, cascade-correlation

The paper is published in Machine Learning, 1994, 16, 59-88. 

Instructions for ftp retrieval of this paper are given below.  If you 
are unable to retrieve and print it and therefore wish to receive a 
hardcopy, please send e-mail to shultz at psych.mcgill.ca

Please do not reply directly to this message.

FTP INSTRUCTIONS:

unix> ftp archive.cis.ohio-state.edu (or 128.146.8.52)
    Name: anonymous
    Password: <your e-mail address>
    ftp> cd pub/neuroprose
    ftp> binary
    ftp> get shultz.balance-ml.ps.Z 
    ftp> quit
unix> uncompress shultz.balance-ml.ps.Z 

Thanks to Jordan Pollack for maintaining this archive. 

Tom Shultz

From cohn at psyche.mit.edu  Fri Aug  5 17:00:03 1994
From: cohn at psyche.mit.edu (David Cohn)
Date: Fri, 5 Aug 94 17:00:03 EDT
Subject: Optimal Experiment Design paper available
Message-ID: <9408052100.AA24701@psyche.mit.edu>

----------------------------------------------------------------------

       Neural Network Exploration Using Optimal Experiment Design
                 AI Lab Memo #1491/CBCL Paper #99
 
                          David A. Cohn
               Dept. of Brain and Cognitive Sciences
	       Massachusetts Institute of Technology
                       Cambridge, MA 02139

I consider the question "How should one act when the only goal is to
learn as much as possible?"  Building on the theoretical results of
Fedorov and MacKay, I apply techniques from Optimal Experiment Design
(OED) to guide the query/action selection of a neural network learner.
I demonstrate that these techniques allow the learner to minimize its
generalization error by exploring its domain efficiently and
completely.  I conclude that, while not a panacea, OED-based
query/action has much to offer, especially in domains where its high
computational costs can be tolerated.
----------------------------------------------------------------------

The above paper is a greatly expanded version of one that appeared at
last year's NIPS, and is available by anonymous ftp to:
             publications.ai.mit.edu

in the file: ai-publications/1994/AIM-1491.ps.Z

It is also available from my home page at:
             http://www.ai.mit.edu/people/cohn/cohn.html

I welcome all comments, questions, and (gentle) criticisms.

 -David Cohn				e-mail: cohn at psyche.mit.edu
  Dept. of Brain & Cognitive Science	phone:  (617) 253-8409
  MIT, E10-243
  Cambridge, MA 02139
  http://www.ai.mit.edu/people/cohn/cohn.html

From harnad at Princeton.EDU  Fri Aug  5 20:25:42 1994
From: harnad at Princeton.EDU (Stevan Harnad)
Date: Fri, 5 Aug 94 20:25:42 EDT
Subject: Motor Control (Feldman): BBS Call for Commentators
Message-ID: <9408060025.AA29660@clarity.Princeton.EDU>

    Below is the abstract of a forthcoming target article by:
                A.G. Feldman & M.F. Levin
                        on:
        POSITIONAL FRAMES OF REFERENCE IN MOTOR CONTROL

This article has been accepted for publication in Behavioral and Brain
Sciences (BBS), an international, interdisciplinary journal providing
Open Peer Commentary on important and controversial current research in
the biobehavioral and cognitive sciences.

Commentators must be current BBS Associates or nominated by a current
BBS Associate. To be considered as a commentator for this article, to
suggest other appropriate commentators, or for information about how to
become a BBS Associate, please send email to:

harnad at clarity.princeton.edu  or harnad at pucc.bitnet        or write to:
BBS, 20 Nassau Street, #240, Princeton NJ 08542  [tel: 609-921-7771]

To help us put together a balanced list of commentators, please give
some indication of the aspects of the topic on which you would bring
your areas of expertise to bear if you were selected as a commentator.
An electronic draft of the full text is available for inspection by
anonymous ftp according to the instructions that follow after the abstract.
____________________________________________________________________

    POSITIONAL FRAMES OF REFERENCE IN MOTOR CONTROL: ORIGIN AND USE

        Anatol G. Feldman (1,2,4) & Mindy F. Levin (2,3,4)
    Institute of Biomedical Engineering, University of Montreal (1)
    Research Centre, Rehabilitation Institute of Montreal, H3S 2J4 (2)
    School of Rehabilitation, University of Montreal (3)
    Centre for Research in Neurological Sciences, University of Montreal (4)
                EMAIL:Feldman at ere.umontreal.ca 

    KEYWORDS: motor control, frames of reference, motoneurons, control
    variables, proprioception, kinaesthesis, equilibrium points,
    multi-muscle systems, pointing, synergy, redundancy problem.

    ABSTRACT: A hypothesis about sensorimotor integration (the lambda
    model) is described and applied to movement control and
    kinesthesia.  The nervous system organizes positional frames of
    reference for the sensorimotor apparatus and produces active
    movements by shifting frames in terms of spatial coordinates.
    Kinematic and electromyographic patterns are not programmed but
    emerge from the dynamic interaction of the system's components,
    including external forces, within the designated frame of
    reference. Motoneuronal threshold properties and proprioceptive
    inputs to motoneurons may be important components in the
    physiological mechanism which produces positional frames of
    reference.  The hypothesis that intentional movements are produced
    by shifting the frame of reference is extended to multi-muscle and
    multi-degrees of freedom systems by providing a solution for the
    redundancy problem the allows the control of a joint alone or in
    combination with other joints to produce any desired limb
    configuration and movement trajectory. For each motor behavior, the
    nervous system uses a strategy which minimizes the number of
    changeable control variables and keep sthe parameters of these
    changes invariant. This is illustrated by examples of simulated
    kinematic and electromyographic signals from single- and
    multi-joint arm movements produced by patterns of control
    variables. Empirical support is provided and additional tests are
    suggested. The model is contrasted with others based on the ideas
    of programming of motoneuronal activity, muscle forces, stiffness
    or movement kinematics.

--------------------------------------------------------------
To help you decide whether you would be an appropriate commentator for
this article, an electronic draft is retrievable by anonymous ftp from
princeton.edu according to the instructions below (the filename is
bbs.feldman). Please do not prepare a commentary on this draft.
Just let us know, after having inspected it, what relevant expertise
you feel you would bring to bear on what aspect of the article.
The file is also retrievable using archie, gopher, and World-Wide Web

URLs (Universal Resource Locators):
ftp://princeton.edu/pub/harnad/BBS/
gopher://gopher.princeton.edu/1ftp%3aprinceton.edu%40/pub/harnad/BBS/
http://192.190.21.10/wic/psych.02.html

-------------------------------------------------------------
   To retrieve a file by ftp from an Internet site, type either:
ftp princeton.edu
   or
ftp 128.112.128.1
   When you are asked for your login, type:
anonymous
   Enter password as queried (your password is your actual userid:
   yourlogin at yourhost.whatever.whatever - be sure to include the "@")
cd /pub/harnad/BBS
   To show the available files, type:
ls
   Next, retrieve the file you want with (for example):
get bbs.feldman
   When you have the file(s) you want, type:
quit
----------
Where the above procedure is not available there are two fileservers:
ftpmail at decwrl.dec.com
       and
bitftp at pucc.bitnet
that will do the transfer for you. To one or the
other of them, send the following one line message:
help
for instructions (which will be similar to the above, but will be in
the form of a series of lines in an email message that ftpmail or
bitftp will then execute for you).

JANET users without ftp can instead utilise the file transfer facilities
at sites uk.ac.ft-relay or uk.ac.nsf.sun.  Full details are available on
request.
-------------------------------------------------------------

From bishopc at helios.aston.ac.uk  Mon Aug  8 10:26:53 1994
From: bishopc at helios.aston.ac.uk (bishopc)
Date: Mon, 8 Aug 1994 14:26:53 +0000
Subject: Lectureships in Neural Computing
Message-ID: <29457.9408081326@sun.aston.ac.uk>



-------------------------------------------------------------------


                     Aston University 

               Neural Computing Research Group

     Department of Computer Science and Applied Mathematics



              LECTURESHIPS IN NEURAL COMPUTING
              --------------------------------


Applications are invited for two lectureships commencing in the 
next academic year. Candidates are expected to have excellent
academic qualifications and a proven record of research.
The appointments will be for an initial period of three years,
with the possibility of subsequent renewal or transfer to a
continuing appointment. 

Successful candidates will be expected to make a substantial
contribution to the research activities of the Department in
the area of neural computing. They will also be expected to 
contribute to the undergraduate and postgraduate teaching 
programmes in computer science.

The Neural Computing Research Group currently comprises three
professors, two lecturers, three postdoctoral research fellows and 
ten postgraduate research students. Current research activity
focusses on principled approaches to neural computing, and spans 
a broad spectrum from theoretical foundations to industrial and
commercial applications. These new appointments will further 
strengthen the research activity of this group.

Salaries will be within the lecturer A and B range 
14,756 to 25,735, and exceptionally up to 28,756 (UK pounds).

If you wish to be considered for one of these positions, please 
send a CV and publications list, together with the names of 3 
referees, to:

    Professor Chris Bishop
    Neural Computing Research Group
    Aston University
    Birmingham B4 7ET, U.K.
    Tel: 021 359 3611 ext. 4270
    Fax: 021 333 6215
    e-mail: c.m.bishop at aston.ac.uk



From bishopc at helios.aston.ac.uk  Mon Aug  8 12:36:57 1994
From: bishopc at helios.aston.ac.uk (bishopc)
Date: Mon, 8 Aug 1994 16:36:57 +0000
Subject: Neural Computing Applications Forum
Message-ID: <992.9408081536@sun.aston.ac.uk>


---------------------------------------------------------------

                     NCAF Two-Day Conference:

     PRACTICAL APPLICATIONS AND TECHNIQUES OF NEURAL NETWORKS

                      14 and 15 September 1994

                 Aston University, Birmingham, UK



14 September 1994
-----------------

Tutorial: Introduction to Neural Networks
(How to get started in Neural Computing)
Richard Palmer, ERA Technology

INVITED TALK: From Regularisation to Cheque Verification
Francoise Fogelman, SLIGOS, France

Workshop: Error Bars and Confidence Limits
(Practical Techniques for Assigning Error Bars 
to Network Predictions)

Odyssey: Stocks, Shakespeare and Sunshine:
(or: Useful Tricks with Radial Basis Functions)
David Lowe, Aston University


Social Event: Canal Boat Trip with Wine and Buffet Dinner



15 September 1994
-----------------

Obstacles and Challenges in the Industrial Application of Neural Networks
Inderjit Sandhu, Barclays Bank PLC

Applying Neural Networks to the Analysis of Small Medical Data Sets
Paul Beatty, University Hospital of South Manchester

The Neural Nose
Gareth Jones, Neotronics Ltd

Hand-printed Character Recognition Using Deformable Models
Chris Williams, Aston University

Determination of Ocean Surface Wind Velocities from Satellite Radar Data
Iain Strachan, AEA Technology

Use of Neural Networks to Derive Helicopter Component Load Spectra
Alf Vella, Cranfield University

Working Together: NCAF and the DTI Neural Computing Programme
Bob Wiggins, DTI

The EPSRC Neural Computing Programme
Catherine Barnes and Peter Bates, EPSRC

A Study of Various Neural Network Techniques for Automotive Systems
M Arain, Lucas Advanced Engineering

Predicting Driver Alertness from Steering Behaviour
Kevin Swingler, Stirling University (with Ford UK)


-------------------------------------------------------------------

                 NEURAL COMPUTING APPLICATIONS FORUM


The Neural Computing Applications Forum (NCAF) was formed in 1990
and has since come to provide the principal mechanism for exchange
of ideas and information between academics and industrialists in 
the UK on all aspects of neural networks and their practical 
applications. NCAF organises four 2-day conferences each year, 
which are attended by around 100 participants. It has its own 
international journal `Neural Computing and Applications' which is 
published quarterly by Springer-Verlag, and it produces a quarterly 
newsletter `Networks'. 

Annual membership rates (Pounds Stirling): 

  Company:     250
  Individual:  140
  Associate     90
  Student:      55

Membership includes free registration at all four annual 
conferences, a subscription to the journal `Neural Computing 
and Applications', and a subscription to `Networks'.
Associate membership is intended for those who are unable to 
attend meetings (eg those overseas) but who wish to 
receive the journal and the newsletter, and does not include
registration at the conferences.

For further information: 
   Tel: +44 (0)784 477271
   Fax: +44 (0)784 472879
   email: c.m.bishop at aston.ac.uk


Chris Bishop

(Chairman, NCAF)


-------------------------------------------------------------------- 

  Professor Chris M Bishop           Tel. +44 (0)21 359 3611 x4270
  Neural Computing Research Group    Fax. +44 (0)21 333 6215
  Dept. of Computer Science          c.m.bishop at aston.ac.uk
  Aston University               
  Birmingham B4 7ET, UK

--------------------------------------------------------------------


From harnad at Princeton.EDU  Mon Aug  8 22:32:16 1994
From: harnad at Princeton.EDU (Stevan Harnad)
Date: Mon, 8 Aug 94 22:32:16 EDT
Subject: Subsymbolic Language Processing: PSYC Multiple Book Review
Message-ID: <9408090232.AA12841@clarity.Princeton.EDU>

                CALL FOR BOOK REVIEWERS

Below is the Precis of SUBSYMBOLIC NATURAL LANGUAGE PROCESSING by
Risto Mikkulainen. This book has been selected for multiple review in
PSYCOLOQUY. If you wish to submit a formal book review (see
Instructions following Precis) please write to psyc at pucc.bitnet
indicating what expertise you would bring to bear on reviewing the book
if you were selected to review it. (If you have never reviewed for
PSYCOLOQUY or Behavioral & Brain Sciences before, it would be helpful
if you could also append a copy of your CV to your message.) If you are
selected as one of the reviewers, you will be sent a copy of the book
directly by the publisher (please let us know if you have a copy
already). Reviews may also be submitted without invitation, but all
reviews will be refereed. The author will reply to all accepted
reviews.

----------------------------------------------------------------------
psycoloquy.94.5.46.language-network.1.miikkulainen  Monday  8 Aug 1994
ISSN 1055-0143 (34 paragraphs, 1 fig, 1 note, 16 references, 609 lines)
PSYCOLOQUY is sponsored by the American Psychological Association (APA)
                Copyright 1994 Risto Miikkulainen

                Precis of:
                SUBSYMBOLIC NATURAL LANGUAGE PROCESSING:
                AN INTEGRATED MODEL OF SCRIPTS, LEXICON, AND MEMORY
                Cambridge, MA: MIT Press, 1993
                15 chapters, 403 Pages

                Risto Miikkulainen
                Department of Computer Sciences
                The University of Texas at Austin
                Austin, TX 78712
                risto at cs.utexas.edu

    ABSTRACT: Distributed neural networks have been very successful in
    modeling isolated cognitive phenomena, but complex high-level
    behavior has been amenable only to symbolic artificial intelligence
    techniques. Aiming to bridge this gap, this book describes
    DISCERN, a complete natural language processing system implemented
    entirely at the subsymbolic level. In DISCERN, distributed neural
    network models of parsing, generating, reasoning, lexical
    processing and episodic memory are integrated into a single system
    that learns to read, paraphrase, and answer questions about
    stereotypical narratives. Using DISCERN as an example, a general
    approach to building high-level cognitive models from distributed
    neural networks is introduced, and the special properties of such
    networks are shown to provide insight into human performance. In
    this approach, connectionist networks are not only plausible models
    of isolated cognitive phenomena, but also sufficient constituents
    for generating complex, high-level behavior.

    KEYWORDS: computational modeling, connectionism, distributed neural
    networks, episodic memory, lexicon, natural language processing,
    scripts.

I. MOTIVATION

1. Recently there has been a great deal of excitement in cognitive
science about the subsymbolic (i.e., parallel distributed processing,
or distributed connectionist, or distributed neural network) approach
to natural language processing. Subsymbolic systems seem to capture a
number of intriguing properties of human-like information processing
such as learning from examples, context sensitivity, generalization,
robustness of behavior, and intuitive reasoning. These properties have
been very difficult to model with traditional, symbolic techniques.

2. Within this new paradigm, the central issues are quite different
from (even incompatible with) the traditional issues in symbolic
cognitive science, and the research has proceeded without much in
common with the past. However, the ultimate goal is still the same: to
understand how human cognition is put together. Even if cognitive
science is being built on a new foundation, as can be argued, many of
the results obtained through symbolic research are still valid, and
could be used as a guide for developing subsymbolic models of cognitive
processes.

3. This is where DISCERN, the computer-simulated neural network model
described in this book (Miikkulainen 1993), fits in. DISCERN is a
purely subsymbolic model, but at the high level it consists of modules
and information structures similar to those of symbolic systems, such
as scripts, lexicon, and episodic memory. At the highest level of
cognitive modeling, the symbolic and subsymbolic paradigms have to
address the same basic issues. Outlining a parallel distributed
approach to those issues is the purpose of DISCERN.

4. In more specific terms, DISCERN aims: (1) to demonstrate that
distributed artificial neural networks can be used to build a
large-scale natural language processing system that performs
approximately at the level of symbolic models; (2) to show that several
cognitive phenomena can be explained at the subsymbolic level using the
special properties of these networks; and (3) to identify central
issues in subsymbolic cognitive modeling and to develop well-motivated
techniques to deal with them. To the extent that DISCERN is successful
in these areas, it constitutes a first step towards subsymbolic natural
language processing.

II. THE SCRIPT PROCESSING TASK

5. Scripts (Schank and Abelson, 1977) are schemas of often-encountered,
stereotypic event sequences, such as visiting a restaurant, traveling
by airplane, and shopping at a supermarket. Each script divides further
into tracks, or established minor variations. A script can be
represented as a causal chain of events with a number of open roles.
Script-based understanding means reading a script-based story,
identifying the proper script and track, and filling its roles with the
constituents of the story. Events and role fillers that were not
mentioned in the story but are part of the script can then be inferred.
Understanding is demonstrated by generating an expanded paraphrase of
the original story, and by answering questions about the story.

6. To see what is involved in the task, let us consider an example of
DISCERN input/output behavior. The following input stories are examples
of the fancy-restaurant, plane-travel, and electronics-shopping
tracks:

    John went to MaMaison. John asked the waiter for lobster. John left
    the waiter a big tip.

    John went to LAX. John checked in for a flight to JFK. The plane
    landed at JFK.

    John went to Radio-Shack. John asked the staff questions
    about CD-players. John chose the best CD-player.

7. DISCERN reads the orthographic word symbols sequentially, one at a
time. An internal representation of each story is formed, where all
inferences are made explicit. These representations are stored in the
episodic memory. The system then answers questions about the stories:

    What did John buy at Radio-Shack?
      John bought a CD-player at Radio-Shack.

    Where did John fly to?
      John flew to JFK.

    What did John eat at MaMaison?
      John ate a good lobster.

With the question as a cue, the appropriate story representation is
retrieved from the episodic memory and the answer is generated word by
word. DISCERN also generates full paraphrases of the input stories.
For example, it generates an expanded version of the restaurant story:

    John went to MaMaison. The waiter seated John. John asked the
    waiter for lobster. John ate a good lobster. John paid the waiter.
    John left a big tip. John left MaMaison.

8. The answers and the paraphrase show that DISCERN has made a number
of inferences beyond the original story. For example, it inferred that
John ate the lobster and the lobster tasted good. The inferences are
not based on specific rules but are statistical and learned from
experience. DISCERN has read a number of similar stories in the past
and the unmentioned events and role bindings have occurred in most
cases. They are assumed immediately and automatically upon reading the
story and have become part of the memory of the story. In a similar
fashion, human readers often confuse what was mentioned in the story
with what was only inferred (Bower et al., 1979; Graesser et al.,
1979).

9. A number of issues can be identified from the above examples.
Specifically, DISCERN has to (1) make statistical, script-based
inferences and account for learning them from experience; (2) store
items in the episodic memory in a single presentation and retrieve them
with a partial cue; (3) develop a meaningful organization for the
episodic memory, based on the stories it reads; (4) represent meanings
of words, sentences, and stories internally; and (5) organize a lexicon
of symbol and concept representations based on examples of how words
are used in the language and form a many-to-many mapping between them.
Script processing constitutes a good framework for studying these
issues, and a good domain for developing an approach towards the goals
outlined above.

III. APPROACH

10. Parallel distributed processing models typically have very little
internal structure. They produce the statistically most likely answer
given the input conditions in a process that is opaque to the external
observer. This is well suited to the modeling of isolated low-level
tasks, such as learning past tense forms of verbs (Rumelhart and
McClelland, 1986) or word pronunciation (Sejnowski and Rosenberg,
1987). Given the success of such models, a possible approach to
higher-level cognitive modeling would be to construct the system from
several submodules that work together to produce the higher-level
behavior.

11. In DISCERN, the immediate goal is to build a complete, integrated
system that performs well in the script processing task. In this sense,
DISCERN is very similar to traditional models in artificial
intelligence. However, DISCERN also aims to show how certain parts of
human cognition could actually be built. The components of DISCERN were
designed as independent cognitive models that can account for
interesting language processing and memory phenomena, many of which are
not even required in the DISCERN task. Combining these models into a
single, working system is one way of validating them. In DISCERN, the
components are not just models of isolated cognitive phenomena; they
are sufficient constituents for generating complex high-level
behavior.

IV. THE DISCERN MODEL

12. DISCERN can be divided into parsing, generating, question
answering, and memory subsystems, each with two modules (figure 1).
Each module is trained in its task separately and in parallel. During
performance, the modules form a network of networks, each feeding its
output to the input of another module.

                       Input text     Output text
                              |         ^
                              V         |
  =================        =================        =================
   Sentence Parser <-------     Lexicon     <------- Sentence Gener.
  =================        =================        =================
     |         |                                           ^     ^
     |         |                                           |     |
     |         +-------+-----------------------+           |     |
     |                 |                       |           |     |
     |                 V                       V           |     |
     |         =================       =================   |     |
     |             Cue Former           Answer Producer ---+     |
     |         =================       =================         |
     |                       |           ^                       |
     |                       |           |                       |
     V                       V           |                       |
  =================        =================        =================
    Story Parser   -------> Episodic Memory -------> Story Generator
  =================        =================        =================

                      Figure 1: The DISCERN Model.

13. The sentence parser reads the input words one at a time and forms
a representation of each sentence. The story parser combines the
sequence of sentences into an internal representation of the story,
which is then stored in the episodic memory. The story generator
receives the internal representation and generates the sentences of the
paraphrase one at a time. The sentence generator outputs the sequence
of words for each sentence. The cue former receives a question
representation, built by the sentence parser, and forms a cue pattern
for the episodic memory, which returns the appropriate story
representation. The answer producer receives the question and the story
and generates an answer representation, which is output word by word by
the sentence generator. The architecture and behavior of each of these
modules in isolation is outlined below.

V. LEXICON

14. The input and output of DISCERN consist of distributed
representations for orthographic word symbols (also called lexical
words). Internally, DISCERN processes semantic concept representations
(semantic words). Both the lexical and semantic words are represented
distributively as vectors of gray-scale values between 0.0 and 1.0. The
lexical representations are based on the visual patterns of characters
that make up the written word; they remain fixed throughout the
training and performance of DISCERN. The semantic representations stand
for distinct meanings and are developed automatically by the system
while it is learning the processing task.

15. The lexicon stores the lexical and semantic representations and
translates between them. It is implemented as two feature maps
(Kohonen, 1989), one lexical and the other semantic. Words whose
lexical forms are similar, such as "LINE" and "LIKE", are represented
by nearby units in the lexical map. In the semantic map, words with
similar semantic content, such as "John" and "Mary", or "Leone's" and
"MaMaison" are mapped near each other. There is a dense set of
associative interconnections between the two maps. A localized activity
pattern representing a word in one map will cause a localized activity
pattern to form in the other map, representing the same word. The
output representation is then obtained from the weight vector of the
most highly active unit. The lexicon thus transforms a lexical input
vector into a semantic output vector and vice versa. Both maps and the
associative connections between them are organized simultaneously,
based on examples of co-occurring symbols and meanings.

16. The lexicon architecture facilitates interesting behavior.
Localized damage to the semantic map results in category-specific
lexical deficits similar to human aphasia (Caramazza, 1988; McCarthy and
Warrington, 1990). For example, the system selectively loses access to
restaurant names, or animate words, when that part of the map is
damaged. Dyslexic performance errors can also be modeled. If the
performance is degraded, for example, by adding noise to the
connections, parsing and generation errors that occur are quite similar
to those observed in human deep dyslexia (Coltheart et al., 1988). For
example, the system may confuse "Leone's" with "MaMaison", or "LINE"
with "LIKE", because they are nearby in the map and share similar
associative connections.

VI. FGREP PROCESSING MODULES

17. Processing in DISCERN is carried out by hierarchically organized
pattern-transformation networks. Each module performs a specific
subtask, such as parsing a sentence or generating an answer to a
question. All these networks have the same basic architecture: they are
three-layer, simple-recurrent backpropagation networks (Elman, 1990),
with the extension called FGREP that allows them to develop distributed
representations for their input/output words.

18. The network learns the processing task by adapting the connection
weights according to the standard on-line backpropagation procedure
(Rumelhart et al., 1986, pp. 327-329). The error signal is propagated
to the input layer, and the current input representations are modified
as if they were an extra layer of weights. The modified representation
vectors are put back in the lexicon, replacing the old representations.
Next time the same words occur in the input or output, their new
representations are used to form the input/output patterns for the
network. In FGREP, therefore, the required mappings change as the
representations evolve, and backpropagation is shooting at a moving
target.

19. The representations that result from this process have a number of
useful properties for cognitive modeling. (1) Since they adapt to the
error signal, they end up coding information most crucial to the task.
Representations for words that are used in similar ways in the examples
become similar. Thus, these profiles of continuous activity values can
be claimed to code the meanings of the words as well. (2) As a result,
the system never has to process very novel input patterns, because
generalization has already been done in the representations. (3) The
representation of a word is determined by all the contexts in which
that word has been encountered; consequently, it is also a
representation of all those contexts. Expectations emerge automatically
and cumulatively from the input word representations. (4) Single
representation components do not usually stand for identifiable
semantic features. Instead, the representation is holographic: word
categories can often be recovered from the values of single
components. (5) Holography makes the system very robust against noise
and damage. Performance degrades approximately linearly as
representation components become defective or inaccurate.

VII. EPISODIC MEMORY

20. The episodic memory in DISCERN consists of a hierarchical pyramid
of feature maps organized according to the taxonomy of script-based
stories. The highest level of the hierarchy is a single feature map
that lays out the different script classes. Beneath each unit of this
map there is another feature map that lays out the tracks within the
particular script. The different role bindings within each track are
separated at the bottom level. The map hierarchy receives a story
representation vector as its input and classifies it as an instance of
a particular script, track, and role binding. The hierarchy thereby
provides a unique memory representation for each script-based story as
the maximally responding units in the feature maps at the three
levels.

21. Whereas the top and the middle level in the hierarchy only serve as
classifiers, selecting the appropriate track and role-binding map for
each input, at the bottom level a permanent trace of the story must
also be created. The role-binding maps are trace feature maps, with
modifiable lateral connections. When the story representation vector is
presented to a role-binding map, a localized activity pattern forms as
a response. Each lateral connection to a unit with higher activity is
made excitatory, while a connection to a unit with lower activity is
made inhibitory. The units within the response now "point" towards the
unit with highest activity, permanently encoding that the story was
mapped at that location.

22. A story is retrieved from the episodic memory by giving it a
partial story representation as a cue. Unless the cue is highly
deficient, the map hierarchy is able to recognize it as an instance of
the correct script and track and form a partial cue for the
role-binding map. The trace feature map mechanism then completes the
role binding. The initial response of the map is again a localized
activity pattern; because the map is topological, it is likely to be
located somewhere near the stored trace. If the cue is close enough,
the lateral connections pull the activity to the center of the stored
trace. The complete story representation is retrieved from the weight
vectors of the maximally responding units at the script, track, and
role-binding levels.

23. Hierarchical feature maps have a number of properties that make
them useful for memory organization: (1) The organization is formed in
an unsupervised manner, extracting it from the input experience of the
system. (2) The resulting order reflects the properties of the data,
the hierarchy corresponding to the levels of variation, and the maps
laying out the similarities at each level. (3) By dividing the data
first into major categories and gradually making finer distinctions
lower in the hierarchy, the most salient components of the input data
are singled out and more resources are allocated for representing them
accurately. (4) Because the representation is based on salient
differences in the data, the classification is very robust, and usually
correct even if the input is noisy or incomplete. (5) Because the
memory is based on classifying the similarities and storing the
differences, retrieval becomes a reconstructive process (Kolodner, 1984;
Williams and Hollan, 1981) similar to human memory.

24. The trace feature map exhibits interesting memory effects that
result from interactions between traces. Later traces capture units
from earlier ones, making later traces more likely to be retrieved.
The extent of the traces determines memory capacity. The smaller the
traces, the more of them will fit in the map, but more accurate cues
are required to retrieve them. If the memory capacity is exceeded,
older traces will be selectively replaced by newer ones. Traces that
are unique, that is, located in a sparse area of the map, are not
affected, no matter how old they are. Similar effects are common in
human long-term memory (Baddeley, 1976; Postman, 1971).

VIII. DISCERN HIGH-LEVEL BEHAVIOR

25. DISCERN is more than just a collection of individual cognitive
models. Interesting behavior results from the interaction of the
components in a complete story-processing system.

26. DISCERN was trained and tested with an artificially generated
corpus of script-based stories consisting of three scripts with three
tracks and three open roles each. The complete DISCERN system performs
very well: at the output, about 98 percent of the words are correct.
This is rather remarkable for a chain of networks that is 9 modules
long and consists of several different types of modules.

27. A modular neural network system can only operate if it is stable,
that is, if small deviations from the normal flow of information are
automatically corrected. It turns out that DISCERN has several built-in
safeguards against minor inaccuracies and noise. The semantic
representations are distributed and redundant, and inaccuracies in the
output of one module are cleaned up by the module that uses the
output. The memory modules clean up by categorical processing: a noisy
input is recognized as a representative of an established class and
replaced by the correct representation of that class. As a result,
small deviations do not throw the system off course, but rather the
system filters out the errors and returns to the normal course of
processing, which is an essential requirement for building robust
cognitive models.

28. DISCERN also demonstrates strong script-based inferencing. Even
when the input story is incomplete, consisting of only a few main
events, DISCERN can usually form an accurate internal representation
of it. DISCERN was trained to form complete story representations
from the first sentence on, and because the stories are stereotypical,
missing sentences have little effect on the parsing process. Once the
story representation has been formed, DISCERN performs as if the script
had been fully instantiated. Questions about missing events and
role-bindings are answered as if they were part of the original story.
If events occurred in an unusual order, they are recalled in the
stereotypical order in the paraphrase. If there is not enough
information to fill a role, the most likely filler is selected and
maintained throughout the paraphrase generation. Such behavior
automatically results from the modular architecture of DISCERN and is
consistent with experimental observations on how people remember
stories of familiar event sequences (Bower et al., 1979; Graesser et
al., 1979).

29. In general, given the information in the question, DISCERN recalls
the story that best matches it in the memory. An interesting issue
is: what happens when DISCERN is asked a question that is inaccurate
or ambiguous, that is, one that does not uniquely specify a story? For
example, DISCERN might have read a story about John eating lobster at
MaMaison, and then about Mary doing the same at Leone's, and the
question could be "Who ate lobster?" Because later traces are more
prominent in the memory, DISCERN is more likely to retrieve the
Mary-at-Leone's story in this case. The earlier story is still in the
memory, but to recall it, more details need to be specified in the
question, such as `Who ate lobster at MaMaison?" Similarly, DISCERN can
robustly retrieve a story even if the question is slightly inaccurate.
When asked "How did John like the steak at MaMaison?", DISCERN
generates the answer "John thought lobster was good at MaMaison",
ignoring the inaccuracy in the question, because the cue is still close
enough to the stored trace. DISCERN does recognize, though, when a
question is too different from anything in the memory, and should not
be answered. For "Who ate at McDonald's?", the cue vector is not close
to any trace, the memory does not settle, and nothing is retrieved.
Note that these mechanisms were not explicitly built into DISCERN,
but they emerge automatically from the physical layout of the
architecture and representations.

IX. DISCUSSION

30. There is an important distinction between scripts (or more
generally, schemas) in symbolic systems, and scripts in subsymbolic
models such as DISCERN. In the symbolic approach, a script is stored
in memory as a separate, exact knowledge structure, coded by the
knowledge engineer. The script has to be instantiated by searching the
schema memory sequentially for a structure that matches the input.
After instantiation, the script is active in the memory and later
inputs are interpreted primarily in terms of this script. Deviations
are easy to recognize and can be taken care of with special
mechanisms.

31. In the subsymbolic approach, schemas are based on statistical
properties of the training examples, extracted automatically during
training. The resulting knowledge structures do not have explicit
representations. For example, a script exists in a neural network only
as statistical correlations coded in the weights. Every input is
automatically matched to every correlation in parallel. There is no
all-or-none instantiation of a particular knowledge structure. The
strongest, most probable correlations will dominate, depending on how
well they match the input, but all of them are simultaneously active at
all times. Regularities that make up scripts can be particularly well
captured by such correlations, making script-based inference a good
domain for the subsymbolic approach. Generalization and graceful
degradation give rise to inferencing that is intuitive, immediate, and
occurs without conscious control, as is script-based inference in
humans. On the other hand, it is very difficult to recognize deviations
from the script and to initiate exception-processing when the automatic
mechanisms fail. Such sequential reasoning would require intervention
of a high-level "conscious" monitor, which has yet to be built in the
connectionist framework.

X. CONCLUSION

32. The main conclusion from DISCERN is that building subsymbolic
models is a feasible approach to understanding mechanisms underlying
natural language processing. DISCERN shows how several cognitive
phenomena may result from subsymbolic mechanisms. Learning word
meanings, script processing, and episodic memory organization are based
on self-organization and gradient-descent in error in this model.
Script-based inferences, expectations, and defaults automatically
result from generalization and graceful degradation. Several types of
performance errors in role binding, episodic memory, and lexical access
emerge from the physical organization of the system. Perhaps most
significantly, DISCERN shows how individual connectionist models can be
combined into a large, integrated system that demonstrates that these
models are sufficient constituents for generating sequential, symbolic,
high-level behavior.

33. Although processing simple script instantiations is a start, there
is a long way to go before subsymbolic models will rival the best
symbolic cognitive models. For example, in story understanding, symbolic
systems have been developed that analyze realistic stories in depth,
based on higher-level knowledge structures such as goals, plans,
themes, affects, beliefs, argument structures, plots, and morals. In
designing subsymbolic models that would do that, we are faced with two
major challenges: (1) how to implement connectionist control of
high-level processing strategies (making it possible to model processes
more sophisticated than a series of reflex responses), and (2) how to
represent and learn abstractions (making it possible to process
information at a higher level than correlations in the raw input
data). Progress in these areas would constitute a major step towards
extending the capabilities of subsymbolic natural language processing
models beyond those of DISCERN.

XI. NOTE

34. Software for the DISCERN system is available through anonymous ftp
from cs.utexas.edu:pub/neural-nets/discern. An X11 graphics demo,
showing DISCERN in processing the example stories discussed in the
book, can be run remotely under the World Wide Web at
http://www.cs.utexas.edu/~risto/discern.html, or by telnet with "telnet
cascais.utexas.edu 30000".

XII. TABLE OF CONTENTS

PART I Overview
1 Introduction
2 Background
3 Overview of DISCERN

PART II Processing Mechanisms
4 Backpropagation Networks
5 Developing Representations in FGREP Modules
6 Building from FGREP Modules

PART III Memory Mechanisms
7 Self-Organizing Feature Maps
8 Episodic Memory Organization: Hierarchical Feature Maps
9 Episodic Memory Storage and Retrieval: Trace Feature Maps
10 Lexicon

PART IV Evaluation
11 Behavior of the Complete Model
12 Discussion
13 Comparison to Related Work
14 Extensions and Future Work
15 Conclusions

APPENDICES
A Story Data
B Implementation Details
C Instructions for Obtaining the DISCERN Software

XIII. REFERENCES

Baddeley, A.D. (1976) The Psychology of Memory. New York: Basic Books.

Bower, G.H., Black, J.B. and Turner, T.J. (1979) Scripts in memory for
text. Cognitive Psychology, 11:177-220.

Caramazza, A. (1988) Some aspects of language processing revealed
through the analysis of acquired aphasia: The lexical system. Annual
Review of Neuroscience, 11:395-421.

Coltheart, M., Patterson, K. and Marshall, J.C., editors (1988) Deep
Dyslexia. London; Boston: Routledge and Kegan Paul. Second edition.

Elman, J.L. (1990) Finding structure in time. Cognitive Science,
14:179-211.

Graesser, A.C., Gordon, S.E. and Sawyer, J.D. (1979) Recognition memory
for typical and atypical actions in scripted activities: Tests for the
script pointer+tag hypothesis. Journal of Verbal Learning and Verbal
Behavior, 18:319-332.

Kohonen, T. (1989) Self-Organization and Associative Memory. Berlin;
Heidelberg; New York: Springer. Third edition.

Kolodner, J.L. (1984) Retrieval and Organizational Strategies in
Conceptual Memory: A Computer Model. Hillsdale, NJ: Erlbaum.

Miikkulainen, R. (1993) Subsymbolic Natural Language Processing:  an
Integrated Model of Scripts, Lexicon, and Memory. Cambridge MA: MIT.

McCarthy, R.A. and Warrington, E.K. (1990) Cognitive Neuropsychology: A
Clinical Introduction. New York: Academic Press.

Postman, L. (1971) Transfer, interference and forgetting. In Kling,
J.W., and Riggs, L.A., editors, Woodworth and Schlosberg's Experimental
Psychology, 1019-1132. New York: Holt, Rinehart and Winston. Third
edition.

Rumelhart, D.E. and McClelland, J.L. (1986) On learning past tenses of
English verbs. In Rumelhart, D.E., and McClelland, J.L., editors,
Parallel Distributed Processing: Explorations in the Microstructure of
Cognition, Volume 2, 216--271. Cambridge, MA: MIT Press.

Rumelhart, D.E., Hinton, G.E. and Williams, R.J. (1986) Learning
internal representations by error propagation. In Rumelhart, D.E. and
McClelland, J.L., editors, Parallel Distributed Processing:
Explorations in the Microstructure of Cognition, Volume 1, 318-362.
Cambridge, MA: MIT Press.

Sejnowski, T. J., and Rosenberg, C. R. (1987) Parallel networks that
learn to pronounce English text. Complex Systems, 1:145--168.

Schank, R.C. and Abelson, R.P. (1977) Scripts, Plans, Goals, and
Understanding: An Inquiry into Human Knowledge Structures. Hillsdale,
NJ: Erlbaum.

Williams, M.D. and Hollan, J.D. (1981) The process of retrieval from
very long-term memory. Cognitive Science, 5:87--119.

--------------------------------------------------------------------

        PSYCOLOQUY Book Review Instructions

The PSYCOLOQUY book review procedure is very similar to the commentary
procedure except that it is the book itself, not a target article, that
is under review. (The Precis summarizing the book is intended to permit
PSYCOLOQUY readers who have not read the book to assess the exchange,
but the reviews should address the book, not primarily the Precis.)

Note that as multiple reviews will be co-appearing, you need only
comment on the aspects of the book relevant to your own specialty and
interests, not necessarily the book in its entirety. Any substantive
comments and criticism -- including points calling for a detailed and
substantive response from the author -- are appropriate. Hence,
investigators who have already reviewed or intend to review this book
elsewhere are still encouraged to submit a PSYCOLOQUY review specifically
written with this specialized multilateral review-and-response feature
in mind.

1.  Before preparing your review, please read carefully
    the Instructions for Authors and Commentators and examine
    recent numbers of PSYCOLOQUY.

2.  Reviews should not exceed 500 lines. Where judged necessary
    by the Editor, reviews will be formally refereed.

3.  Please provide a title for your review. As many
    commentators will address the same general topic, your
    title should be a distinctive one that reflects the gist
    of your specific contribution and is suitable for the
    kind of keyword indexing used in modern bibliographic
    retrieval systems. Each review should also have a brief
    (~50-60 word) Abstract

4.  All paragraphs should be numbered consecutively. Line length
    should not exceed 72 characters.  The review should begin with
    the title, your name and full institutional address (including zip
    code) and email address.  References must be prepared in accordance
    with the examples given in the Instructions.  Please read the
    sections of the Instruction for Authors concerning style,

    INSTRUCTIONS FOR PSYCOLOQUY AUTHORS AND COMMENTATORS

PSYCOLOQUY is a refereed electronic journal (ISSN 1055-0143) sponsored
on an experimental basis by the American Psychological Association
and currently estimated to reach a readership of 40,000. PSYCOLOQUY
publishes brief reports of new ideas and findings on which the author
wishes to solicit rapid peer feedback, international and
interdisciplinary ("Scholarly Skywriting"), in all areas of psychology
and its related fields (biobehavioral science, cognitive science,
neuroscience, social science, etc.). All contributions are refereed.

Target article length should normally not exceed 500 lines [c. 4500 words].
Commentaries and responses should not exceed 200 lines [c. 1800 words].

All target articles, commentaries and responses must have (1) a short
abstract (up to 100 words for target articles, shorter for commentaries
and responses), (2) an indexable title, (3) the authors' full name(s)
and institutional address(es).

In addition, for target articles only: (4) 6-8 indexable keywords,
(5) a separate statement of the authors' rationale for soliciting
commentary (e.g., why would commentary be useful and of interest to the
field? what kind of commentary do you expect to elicit?) and
(6) a list of potential commentators (with their email addresses).

All paragraphs should be numbered in articles, commentaries and
responses (see format of already published articles in the PSYCOLOQUY
archive; line length should be < 80 characters, no hyphenation).

It is strongly recommended that all figures be designed so as to be
screen-readable ascii. If this is not possible, the provisional
solution is the less desirable hybrid one of submitting them as
postscript files (or in some other universally available format) to be
printed out locally by readers to supplement the screen-readable text
of the article.

PSYCOLOQUY also publishes multiple reviews of books in any of the above
fields; these should normally be the same length as commentaries, but
longer reviews will be considered as well. Book authors should submit a
500-line self-contained Precis of their book, in the format of a target
article; if accepted, this will be published in PSYCOLOQUY together
with a formal Call for Reviews (of the book, not the Precis). The
author's publisher must agree in advance to furnish review copies to the
reviewers selected.

Authors of accepted manuscripts assign to PSYCOLOQUY the right to
publish and distribute their text electronically and to archive and
make it permanently retrievable electronically, but they retain the
copyright, and after it has appeared in PSYCOLOQUY authors may
republish their text in any way they wish -- electronic or print -- as
long as they clearly acknowledge PSYCOLOQUY as its original locus of
publication. However, except in very special cases, agreed upon in
advance, contributions that have already been published or are being
considered for publication elsewhere are not eligible to be considered
for publication in PSYCOLOQUY,

Please submit all material to psyc at pucc.bitnet or psyc at pucc.princeton.edu
Anonymous ftp archive is DIRECTORY pub/harnad/Psycoloquy HOST princeton.edu


From fellous at selforg.usc.edu  Tue Aug  9 19:05:48 1994
From: fellous at selforg.usc.edu (Jean-Marc Fellous)
Date: Tue, 9 Aug 1994 16:05:48 -0700
Subject: Position in Bochum
Message-ID: <199408092305.QAA08804@selforg.usc.edu>



Please post the following Job announcement:


-----------------------

The following announcement is related to a C3 assistant professor at the
University of Bochum (Germany) in the field of neuro-computation.
It is intentionally kept in its original German form, since knowledge of German
is prefered but not required (However, non-German speaking applicants are expected
to eventually learn German ...).



              Ruhr-Universit"at Bochum

Am Institut f"ur Neuroinformatik ist eine C3-Professur
"Neuroinformatik" zu besetzen.

Das Institut ist eine zentrale wissenschaftliche Einrichtung der
Universit"at mit den Abteilungen Systembiophysik und Theoretische
Biologie.  Arbeitsschwerpunkte sind Prinzipien der Selbst-organisation
und Informationsverarbeitung in neuronaler Architektur.  Die Position
schlie{\ss}t die Mitwirkung der kollegialen Leitung des Instituts ein.
Zu den Lehraufgaben geh"oren Vorlesungen "uber Neuronale Netze sowie
"uber technisch nutzbare Organisationsprinzipien biologischer
neuronaler Systeme.

Es wird erwartet, da{\ss} die zu berufende Pers"onlichkeit  
mindestens in einem der folgenden Gebiete wissenschaftlich  
ausgewiesen ist:
	- Analyse und Anwendung biologischer neuronaler 
          Organisationsprinzipien
	- K"unstliche Neuronale Netze
	- Entwurf von Systemen in neuronaler Architektur
	- Probleme der Selbstorganisation

Neben Erfahrungen im theoretischen Bereich (Systemtheorie,  
nicht-lineare Dynamik) wird Interesse an  
anwendungsorientierten Problemen vorausgesetzt. Eine  
Kooperation mit dem Zentrum f"ur Neuroinformatik in Bochum,  
das bevorzugt anwendungsorientierte Probleme bearbeitet,  
ist m"oglich.

Die Ruhr-Universit"at Bochum bem"uht sich um die F"orderung von Frauen
in Forschung und Lehre.  Schwerbehinderte werden bei gleicher
Qualifikation bevorzugt.

Ihre schriftliche Bewerbung mit den "ublichen Unterlagen  
richten Sie bitte an den Rektor der Ruhr-Universit"at  
Bochum, 44780 Bochum, Germany.



----- End Included Message -----


From druck at afit.af.mil  Wed Aug 10 15:40:12 1994
From: druck at afit.af.mil (Dennis W. Ruck)
Date: Wed, 10 Aug 94 15:40:12 -0400
Subject: CFP: SPIE Applications and Science of Artificial Neural Networks VI
Message-ID: <9408101940.AA06900@gandalf.afit.af.mil>

                        CALL FOR PAPERS

                       SPIE Conference on
  Applications and Science of Artificial Neural Networks VI
 	     

	                Orlando, Florida
		        17-21 April 1995    


You are invited to submit a paper to the SPIE conference Applications and  
Science of Artificial Neural Networks VI. This conference will be held in  
conjunction with nearly 40 other conferences on topics in object recognition,  
aerospace sensing, and photonics. See below for a complete list. 


--------------------------------------------------------------
              Announcement and Call for Papers
	                   for
  Applications and Science of Artificial Neural Networks VI
--------------------------------------------------------------
Conference Chairs: Steven K. Rogers, Dennis W. Ruck, Air Force
Institute of Technology

Program Committee: Stanley C. Ahalt, The Ohio State Univ.; James C.
Bezdek, Univ. of West Florida; Joe R. Brown, Microelectronics and
Computer Technology Corp.; Lee A. Feldkamp, Ford Motor Co.; Michael
Georgiopoulos, Univ. of Central Florida; Joydeep Ghosh, Univ. of
Texas/Austin; Charles W. Glover, Oak Ridge National Lab.; John B.
Hampshire, II, Jet Propulsion Lab.; Richard P. Lippmann, MIT
Lincoln Lab.; Harley R. Myler, Univ. of Central Florida; Mary Lou
Padgett, Auburn Univ.; Kevin L. Priddy, Accurate Automation Corp.;
Gintaras V. Puskorius, Ford Motor Co.;
Donald F. Specht, Lockheed Palo Alto Research Lab.; Gregory L.
Tarr, Air Force Phillips Lab.; Gary Whittington, Univ. of Aberdeen
(UK); Rodney G. Winter, Dept. of Defense

The focus of this conference is on real-world applications of artificial neural  
networks and on recent theoretical developments applicable to current  
applications. The goal of this conference is to provide a forum for interaction  
between researchers and industrial/government agencies with information  
processing requirements. Papers that investigate advantages/disadvantages of  
artificial neural networks in specific real-world applications will be  
presented. Papers that clearly state existing problems in information  
processing that could potentially be solved by artificial neural networks will  
also be considered.

Sessions will concentrate on:

* innovative applications of artificial neural networks to solve
  real-world problems
* comparative performance in applications of target recognition,
  object recognition, speech processing, speaker identification,
  cochannel processing, signal processing in realistic
  environments, robotics, process control, and image processing
* demonstrations of properties and limitations of existing or new
  artificial neural networks as shown by or related to an
  application
* environments for artificial neural networks development and
  implementation with specific applications used to demonstrate
  features of the systems
* hardware implementation technologies that are general purpose or
  application specific
* knowledge acquisition and representation
* biologically inspired visual representation techniques
* decision support systems
* artificial life
* cognitive science
* hybrid systems (fuzzy, neural, genetic)
* neurobiology
* optimization
* sensation and perception
* system identification
* financial applications
* time series analysis and prediction
* pattern recognition
* medical applications
* intelligent control
* robotics.

------------------------------------------------------------------
                    List of Conferences
------------------------------------------------------------------

The following conferences will all be held in Orlando, Florida 17-21 April 1995  
at the Marriott Orlando World Center:

  1.  Public Safety/Law Enforcement Technology
  2.  Photonics for Space Environments III
  3.  Space Environmental, Legal, and Safety Issues
  4.  Imaging Spectrometry
  5.  Commercialization of High-Resolution Satellite Imagery
      for Dual-Use Applications
  6.  Infrared Detectors and Instrumentation for Astronomy
  7.  Spaceborne Interferometry II
  8.  Space Telescopes and Instruments III
  9.  Fiber Optics in Astronomical Applictions
 10.  Telescope Control Systems
 11.  Distributed Interactive Simulation (Critical Reviews)
 12.  Helmet- and Head-Mounted Displays and Symbology Design
      Requirements II
 13.  Cockpit Displays II
 14.  Space Guidance, Control, and Tracking II
 15.  Synthetic Vision for Vehicle Guidance and Control
 16.  Acquisition, Tracking, and Pointing IX
 17.  Applied Laser Radar Technology II
 18.  Air Traffic Control Technologies
 19.  Technologies for Advanced Land Combat (Critical Reviews)
      Part I: Rapid Force Projection Initiative
      Part II: Advanced Vehicle Technologies
      Part III: Information Sciences for Digitizing the Battlefield
 20.  Detection Technologies for Mines and Minelike Targets
 21.  Targets and Backgrounds: Characterization and Representation
 22.  Atmospheric Propagation and Remote Sensing IV
 23.  Tactical Control Technologies
 24.  Test and Evaluation of Defense-Related Infrared Detectors
      and Arrays
 25.  Infrared Imaging Systems: Design, Analysis, Modeling,
      and Testing VI
 26.  Signal Processing, Sensor Fusion, and Target Recognition IV
 27.  Algorithms for Synthetic Aperture Radar Imagery II
 28.  Integration Photogrammetric Techniques with Scene Analysis
      and Machine Vision II
 29.  Automatic Object Recognition V
 30.  Smart Infrared Focal Plane Arrays and Technology
 31.  Transition of Optical Processors into Systems 1995
 32.  Optical Pattern Recognition VI
 33.  Visual Information Processing IV
 34.  Applications and Science of Artificial Neural Networks VI
 35.  Applications of Fuzzy Logic Technology II
 36.  Thermosense XVII: An International Conference on Thermal
      Sensing and Imaging Diagnostic Applications
 37.  Flat Panel Displays for Defense Applications
      (Critical Reviews)
 38.  Digital Signal Processing Technology (Critical Reviews)


------------------------------------------------------------------
                    General Information
------------------------------------------------------------------
SPIE's International Symposium on
Aerospace/Defense Sensing and Dual-Use Photonics
17-21 April 1995
Marriott's Orlando World Center
Resort and Convention Center
Orlando, Florida USA

SPIE's 1995 Aerospace/Defense Sensing and Dual-Use Photonics
Symposium will be held at:
Marriott's Orlando World Center Hotel
8701 World Center Drive
Orlando, Florida 32821-6398
Phone: 407/239-4200 or 800/621-0638 (outside Florida)
Fax: 407/239-5958

Accommodations
--------------
SPIE will reserve a block of rooms for attendees at the Marriott
Orlando World Center Hotel. Room rates at the Marriott will be $129
single and $142 double plus tax. Alternate hotels in the immediate
area will also be available. Information concerning hotels and
prices will be announced in the advance program.


Advance Technical Program
-------------------------
The comprehensive Advance Technical Program for this symposium will
list conferences, paper titles and authors in order of
presentation, educational short courses schedule including course
descriptions and instructor biographies, and an outline of all
planned special events. Call SPIE at 206/676-3290 (Pacific Time) to
request that a copy be sent to you when it becomes available in
January 1995.


Conference Registration
-----------------------
The following registration fees for SPIE's International Symposium
on Aerospace/Defense Sensing and Dual-Use Photonics are preliminary
and included to assist you in planning.

Conference Fees without Proceedings       Member      Nonmember
  Attendee Full Conference.................$360.........$420
  One day...................................160..........190
  Author Full Conference*...................325..........385
  Author One Day*...........................160..........190
  Student....................................85...........95

* Author fee includes a proceedings

Short Course Fees

                              Member    Nonmember
Half-day course (3.5 hr).......$145.......$170
Full-day course (6.5 hr)........265........310
Two-day course (12 hr)..........485........570

Florida sales tax will be added to short course fees.


How to Contact SPIE
-------------------
If you have further questions, or need assistance, please send a
message to info-optolink-service at mom.spie.org. You will receive a
response from an SPIE staff member.


Join SPIE Today
---------------
Keep in touch with the dynamic world of optics and optoelectronics
by becoming a member of SPIE.

Full SPIE Membership
Joining SPIE as a full member provides you with many benefits,
including:

* Voting privileges
* Eligibility to hold SPIE office.
* Subscription to OE Reports
* Subscription to Optical Engineering, SPIE's monthly journal
* Annual SPIE Member Guide
* Full member discounts (~20%) on SPIE publications
* Full member discounts (~15%) on SPIE conferences and short
  courses
* Discounts on publications from other publishers as available
* Member rates for SPIE-cosponsored technical events

$85 in North America/$95 outside North America (Student Memberships
and Associate Student Memberships available at reduced rates.)


Working Group Membership
------------------------
Working Groups are interactive networks that foster professional
contacts and information flow among technically related
individuals, groups, companies, and institutions in specific areas
of technology.

Individual Membership ($15) Group Memberships and Corporate
Memberships are available. Contact SPIE for complete list of
working groups and benefits.

-------------------------------------------------------------------

              Abstract Due Date:  19 September 1994
    On-Site Proceedings Manuscript Due Date:  23 January 1995

       Manuscript due date for on-site proceedings must be
                      strictly observed.

------------------------------------------------------------------
For a complete text of the Announcement and Call for Papers for
SPIE's International Symposium on Aerospace/Defense Sensing and
Dual-Use Photonics, contact SPIE at either the European Office, or
International Headquarters addresses below.

Contact addresses:
SPIE in Europe:
  SPIE European Office
  c/o HIB-INFONET
  P.O. Box 4463
  N-5028 Bergen, Norway
  Phone: 47 55 54 37 84
  Fax: 47 55 96 21 75
  E-mail: spie at hibinc.no

SPIE International Headquarters
  P.O. Box 10
  Bellingham, WA  98227-0010 USA
  Phone: 206/676-3290
  Fax:  206/647-1445
  E-mail: spie at spie.org
  Telnet/FTP: spie.org
  World Wide Web URL: http://www.spie.org

-------------------------------------------------------------------
SPIE--The International Society for Optical Engineering

SPIE is a nonprofit society dedicated to advancing engineering and
scientific applications of optical, electro-optical, and
optoelectronic instrumentation, systems, and technology. Its
members are scientists, engineers, and users interested in the
reduction to practice of these technologies. SPIE provides the
means for communicating new developments and applications to the
scientific, engineering, and user communities through its
publications, symposia, and short courses.

SPIE is dedicated to bringing you quality electronic media and
online services.
-------------------------------------------------------------------

Dennis W. Ruck               Air Force Institute of Technology
d.ruck at ieee.org              Wright-Patterson AFB, Ohio
AFIT/ENG, Bldg 642, 2950 P ST, Wright-Patterson AFB OH 45433-7765
Ph. (513) 255-6565 ext. 4285   Fax (513) 476-4055


From UBJTP69 at CCS.BBK.AC.UK  Thu Aug 11 14:41:00 1994
From: UBJTP69 at CCS.BBK.AC.UK (Gareth)
Date: Thu, 11 Aug 94 14:41 BST
Subject: Thesis printing problem
Message-ID: <mailman.698.1149540293.24850.connectionists@cs.cmu.edu>

It has been reported that the file gaskell.thesis.ps.Z in the neuroprose
archive will not print out from some unix systems.  If this is the case,
edit the file and delete the initial character (^D).  The file should
then print out properly.

Sorry for any inconvenience.

Gareth Gaskell

From pluto at cs.ucsd.edu  Thu Aug 11 16:49:04 1994
From: pluto at cs.ucsd.edu (Mark Plutowski)
Date: Thu, 11 Aug 94 13:49:04 -0700
Subject: Paper in Neuroprose: estimating generalization with cross-validation
Message-ID: <9408112049.AA14173@beowulf>

FTP-host: archive.cis.ohio-state.edu
FTP-filename: /pub/neuroprose/pluto.imse.ps.Z

|This is related to the generalization debate on Machine Learning List.|


By popular request, the following paper has been placed in Neuroprose.

Title: 	   "Cross-validation estimates integrated mean squared error." 

Authors:   Plutowski , M. (1,3), S. Sakata (2), H. White (2,3). 
	   (1) Computer Science and Engineering
	   (2) Economics
	   (3) Institute for Neural Computation
	   (All at UCSD).


A discussion on the Machine Learning List prompted the question 

"Have theoretical conditions been established under which 
cross-validation is justified?"  

The answer is "Yes."  The statistical literature abounds with
application-specific and model-specific demonstrations that
cross-validation is statistically accurate and precise for use as 
a real-world estimate of an ideal measure of generalization known
as Integrated Mean Squared Error (IMSE).  IMSE is the
average mean squared error, averaged over all training sets of 
a particular size.   IMSE is closely related to Prediction Risk,
(aka statistical risk) therefore such results are applicable to 
statistical risk as well (as averaged over training sets of a
particular size).  See Plutowski's thesis in Neuroprose/Thesis for
explicit relationship between IMSE and statistical risk.

This paper extends such results to apply to nonlinear regression 
in general.  Strong convergence (w.p.1) and unbiasedness are proved.

The key assumption is that the training and test data be
independent and identically distributed (i.i.d.) - therefore, 
data must be drawn from the same space (stationarity), 
independent of previously sampled datum.

Note that if training data are explicitly excluded from the
test sample, then the i.i.d. assumption does not hold,
since in this case the test sample is drawn from a space that 
depends upon (is conditioned by) particular choice of training sample.  

Therefore, the measure of generalization referred to in the raging 
debate on the Machine Learning List would not meet the conditions 
employed by the results in this paper.


Filename:  pluto.imse.ps.Z.  
Title: 	   "Cross-validation estimates integrated mean squared error." 
File size: 97K compressed, 242K uncompressed.  
	   17 single-spaced pages (8 pages of text, the remainder is a
	   mathematical appendix).

Email contact: pluto at cs.ucsd.edu.  

SUBJECT: theorems proving cross-validation is a statistically 
accurate and precise estimator of an ideal measure of generalization.
Abridged version of this appeared in NIPS 6.


= Mark Plutowski


PS:  Sorry, no hard copies available.


From giles at research.nj.nec.com  Thu Aug 11 19:01:15 1994
From: giles at research.nj.nec.com (Lee Giles)
Date: Thu, 11 Aug 94 19:01:15 EDT
Subject: Corrected ps file of previously announced TR.
Message-ID: <9408112301.AA10157@fuzzy>


It seems that a figure in the postscript file of this TR 
generated printing problems for certain printers. We changed the figure 
to eliminate this problem. We apologize to anyone who was inconvenienced. 
In the revised TR, we also included some references that initially were 
unintentionally excluded.

Lee Giles, Bill Horne, Tsungnan Lin


_________________________________________________________________________________________



                 Learning a Class of Large Finite State Machines
                      with a Recurrent Neural Network

        UNIVERSITY OF MARYLAND TECHNICAL REPORT UMIACS-TR-94-94 AND CS-TR-3328

                 C. L. Giles[1,2], B. G. Horne[1], T. Lin[1,3] 

      [1] NEC Research Institute, 4 Independence Way, Princeton, NJ  08540
            [2]UMIACS, University of Maryland, College Park, MD 20742
          [3] EE Department, Princeton University, Princeton, NJ 08540

                      {giles,horne,lin}@research.nj.nec.com
                
One of the issues in any learning model is how it scales with problem size. Neural 
networks have not been immune to scaling issues. We show that a dynamically-
driven discrete-time recurrent network (DRNN) can learn rather large grammatical 
inference problems when the strings of a finite memory machine (FMM) are 
encoded as temporal sequences. FMMs are a subclass of finite state machines 
which have a finite memory or a finite order of inputs and outputs. The DRNN that 
learns the FMM is a neural network that maps directly from the sequential 
machine implementation of the FMM. It has feedback only from the output and not 
from any hidden units; an example is the recurrent network of Narendra and 
Parthasarathy. (FMMs that have zero order in the feedback of outputs are called 
definite memory machines and are analogous to Time-delay or Finite Impulse 
Response neural networks.) Due to their topology these DRNNs are as least as 
powerful as any sequential machine implementation of a FMM and should be 
capable of representing any FMM. We choose to learn ``particular FMMs.\' Specif
ically, these FMMs have a large number of states (simulations are for $256$ and 
$512$ state FMMs) but have minimal order, relatively small depth and little logic 
when the FMM is implemented as a sequential machine. Simulations for the num
ber of training examples versus generalization performance and FMM extraction 
size show that the number of training samples necessary for perfect generalization 
is less than that sufficient to completely characterize the FMM to be learned. This 
is in a sense a best case learning problem since any arbitrarily chosen FMM with a 
minimal number of states would have much more order and string depth and most 
likely require more logic in its sequential machine implementation.


--------------------------------------------------------------------------------------

                          FTP INSTRUCTIONS

                unix> ftp cs.umd.edu (128.8.128.8)
                Name: anonymous
                Password: (your_userid at your_site)
                ftp> cd pub/papers/TRs
                ftp> binary
                ftp> get 3328.ps.Z
                ftp> quit
                unix> uncompress 3328.ps.Z

                                 OR

                unix> ftp external.nj.nec.com (138.15.10.100)
                Name: anonymous
                Password: (your_userid at your_site)
                ftp> cd pub/giles/papers
                ftp> binary
                ftp> get large.fsm.ps.Z
                ftp> quit
                unix> uncompress large.fsm.ps.Z

--------------------------------------------------------------------------------------


--                                 
C. Lee Giles / NEC Research Institute / 4 Independence Way
Princeton, NJ 08540 / 609-951-2642 / Fax 2482
==




From grino at ic.upc.es  Fri Aug 12 16:11:50 1994
From: grino at ic.upc.es (Robert Grino)
Date: Fri, 12 Aug 1994 16:11:50 UTC+0100
Subject: Paper in Neuroprose: Nonlinear System Identification Using Additive
 Dynamic Neural Networks
Message-ID: <232*/S=grino/OU=ic/O=upc/PRMD=iris/ADMD=mensatex/C=es/@MHS>

FTP-host: archive.cis.ohio-state.edu
FTP-filename: /pub/neuroprose/grino.sysid.ps.Z

The file grino.sysid.ps.Z is now available for
copying from the Neuroprose repository:

NONLINEAR SYSTEM IDENTIFICATION USING ADDITIVE DYNAMIC NEURAL NETWORKS
R. Grino (grino at ic.upc.es)
Instituto de Cibernetica - ESAII
Univ. Politecnica Catalunya
Diagonal, 647, 2nd floor
08028-Barcelona, SPAIN
(Reprint of a SICICA'94 (IFAC) paper)


ABSTRACT:
In this work additive dynamic neural models are used for the 
identification of nonlinear plants in on-line operation. In order to accomplish
this task a gradient parameter adaptation method based in sensitivity 
analysis is formulated taking into account that the parameters of the model 
are arranged in matrix form. This methodology is applied to several nonlinear 
systems in simulation and with a real dataset to verify its performance.

=============================================================================
   Robert Grino                   E-mail: grino at ic.upc.es
   Instituto de Cibernetica               
   Diagonal,647, 2nd floor        FAX number: (343) 4016605
   08028 - Barcelona              
       SPAIN    
=============================================================================

From dhw at santafe.edu  Fri Aug 12 16:32:37 1994
From: dhw at santafe.edu (dhw@santafe.edu)
Date: Fri, 12 Aug 94 14:32:37 MDT
Subject: No subject
Message-ID: <9408122032.AA06773@chimayo.santafe.edu>

Mark Plutowski recently said on connectionist:


>>>
A discussion on the Machine Learning List prompted the question 

"Have theoretical conditions been established under which 
cross-validation is justified?"  

The answer is "Yes." 
>>>



Mark is being polite by not using names; I am the one he is
(implicitly) taking to task, for the following comment on the
ML list:

>>>
. an assumption *must always* be
present if we are to have any belief in learnability in the problem
at hand...
However, to give just one example, nobody has yet delineated
just what those assumptions are for the technique of cross-validation.
>>>

Mark is completely correct in his (implicit) criticism.

As he says, there has in fact been decades of work analyzing 
cross-validation from a sampling theory perspective. Mark's thesis is 
a major contribution to this literature. Any implications coming from my
message that such literature doesn't exist or is somehow invalid
are completely mistaken and were not intended. (Indeed, I've had 
several very illuminating discussions with Mark about his thesis!)

The only defense for the imprecision of my comment is that I made 
it in the context of the ongoing discussion on the ML list that 
Mark referred to. That discussion concerned off-training set error 
rather than iid error, so my comments implicitly assumed off-training 
set error. And as Mark notes, his results (and the others in the 
literature) don't extend to that kind of error.

(Another important distinction between the framework Mark uses and the
implicit framework in the ML list discussion is that the latter has
been concerned w/ zero-one loss, whereas Mark's work concentrates on
quadratic loss. The no-free-lunch results being discussed in the ML
list change form drastically if one uses quadratic rather than
zero-one loss. That should be no surprise, given, for example, the
results in Michael Perrone's work.)

While on the subject of cross-validation and iid error though, it's
interesting to note that there is still much to be understood. For
example, in the average-data scenario of sampling theory statistics
that Mark uses, asymptotic properties are better understood than
finite data properties. And in the Bayesian this-data framework, very
little is known for any data-size regime (though some work by Dawid on
this subject comes to mind).




David Wolpert




From mccauley at ecn.purdue.edu  Thu Aug 11 21:51:21 1994
From: mccauley at ecn.purdue.edu (James Darrell McCauley)
Date: Thu, 11 Aug 1994 20:51:21 -0500
Subject: connectionists suggestion
In-Reply-To: <9408112049.AA14173@beowulf>
Message-ID: <199408120151.UAA03022@alx3.ecn.purdue.edu>

Mark Plutowski (pluto at cs.ucsd.edu) writes on 11 Aug 94:
>FTP-host: archive.cis.ohio-state.edu
>FTP-filename: /pub/neuroprose/pluto.imse.ps.Z

May I suggest that posters be encouraged to use
the URL format and *not* post instructions on the
usage of ftp? I.e., the following would be sufficient:

ftp://archive.cis.ohio-state.edu/pub/neuroprose/pluto.imse.ps.Z

(this way readers may just cut/paste the string into xmosaic
or whatever browser they use).

Not picking on Dr Plutoski, just a suggestion that came to mind...

--Darrell McCauley, aka jdm5548 at diamond.tamu.edu

James Darrell McCauley, Purdue Univ, West Lafayette, IN 47907-1146, USA
mccauley at ecn.purdue.edu, mccauley%ecn at purccvm.bitnet, pur-ee!mccauley

From harnad at Princeton.EDU  Fri Aug 12 23:53:54 1994
From: harnad at Princeton.EDU (Stevan Harnad)
Date: Fri, 12 Aug 94 23:53:54 EDT
Subject: French Cognitive Science Conference
Message-ID: <9408130353.AA25579@clarity.Princeton.EDU>

From: payette at uranus.atoci.uqam.ca
Subject: French International Cognitive Science Conference

                              ANNOUNCEMENT
        The Seventh Colloquium of the Jacques Cartier Center
                              Lyon, France. 

            THE COGNITIVE SCIENCES: FROM COMPUTATIONAL MODELS  
                       TO THE  PHILOSOPHY OF MIND                        
               
                           under the aegis of:
            the Pole Rhones-Alpes of the Cognitive Sciences, 
      Programme Interdisciplinaire de Recherche Cognisciences,CNRS
                    Universite du Quebec a Montreal 
                        Universite de Montreal 
                       Universite Joseph Fourier
                       Universite Claude Bernard

Scientific committee:
Denis Fisette  (Universite du Quebec a Montreal, Quebec)
Marc Jeannerod (Universite Claude Bernard, Lyon)
Daniel Laurier (Universite de Montreal, Quebec)
Daniel Payette (Universite du Quebec a Montreal, Quebec)
Vincent Rialle (Universite Joseph Fourier, Grenoble)
Guy Tiberghien (Universite Pierre Mendes-France, Grenoble)

Coordination in North America: Daniel Payette and Denis Fisette  
Universite du Quebec a Montreal, Dpt de Philosophie, Dpt Psychology; C.P.
8888,Succ A, Montreal (Quebec) H3C-3P8, Canada; 
E.mail : payette at uranus.atoci.uqam.ca; 
tel (+514) 987 8418;  Fax: (+514) 9876721 

Coordination in Europe: Vincent Rialle
Universite J. Fourier, Labo.TIMC-IMAG, Faculte de Medecine, 38706 LaTronche
Cedex  E.mail: Vincent.Rialle at imag.fr; 
Tel. (+33) 76 63 71 87;  Fax. (+33) 76 51 8667

DATES: Wednesday, November 30th to Friday, December 2nd 1994

CONFERENCE SITE:  
Amphitheatre CHARLES BERAUDIER. Conseil Regional RHONE-ALPES,78 route de
Paris 69751 CHARBONNIERES-les-BAINS. France

*Talks will only be given by invited speakers. (Simultaneous
French-English and English-French will be provided).
                         
                THEME OF COLLOQUIUM

The modeling of mental processes in the various human cognitive
activities has generated increasing interest in the scientific world
today. Cognitive models, cognitive simulations, auto-organization,
adaptation, emergence, genetic selection, Darwinian mentalism and
enaction are active research topics in neurological and psychological
theory.

The cognitive sciences offer a continuum of research extending from the
engineering sciences to the philosophy of mind, including the
neurosciences, cognitive psychology, linguistics, semantics, semiotics
and artificial intelligence. Three subconferences will organize
themselves around the following major complementary themes: (i)
Modeling (cognitive and brain functions), (ii) Philosophy of Mind and
Epistemology, and (iii) Applications (AI, technical and computational
engineering).

(i) Modeling is a point of intersection for all these specialties
because it includes the modeling of functions and dysfunctions of the
central nervous system, the neurocomputer sciences, the modeling of
psychocognitive and mental processes, the emergence of intentional
structure on the basis of biological structure, enaction, genetic
algorithms, neural networks, artificial "life," etc.

(ii) The philosophical and epistemological subcomponent poses questions
like the following:  Can we elaborate mathematical models of the mind
and use them to describe and explain human behavior? Are we aiming
toward a mathematical model of the mind? Can we capture the formal
principles of the development and emergence of cognition? Can we
technologically recreate thought? Is the computational symbolic
paradigm, which has imposed itself for the last decades, still a
powerful conceptual tool or is it proving too reductionistic and if so,
how? What is the epistemological status of, for example, the
alternative proposed by the parallel distributed model to the
computational models of classical cognitivism? What is the relation os
the modeling activity of the cognitive and neurosciences and human
experience?

(iii) The applications subconference will consider practical domains in
which scientific results have been applied in the treatment of
language, the automated cognitive analyses of textual documents (an
intersection of linguistics, semantics, semiotics and artificial
intelligence), aids to decision making, applications in sensory
information processing, etc.

PREPROGRAM

WEDNESDAY, 30 November 1994
8h15 - 8h30
Allocution d'accueil du Conseil Regional
8h30 - 9h
Guy Tiberghien (Universite Pierre Mendes-France, Grenoble)
Introduction

SESSION 1 : Modelisation neuro et psycho-cognitives

9h - 9h-30
Jean Francois Le Ny (Universite Paris-Sud, psychologie cognitive)
Pourquoi les modeles cognitifs devraient-ils etre calculatoires ?
9h30 - 9h45 Discussion
9h-45 - 10h15
Marc Jeannerod (Universite Claude Bernard, Lyon, neurosciences)
Le cerveau representationnel
10h15 - 10h30 Discussion
10h30 - 10h45 PAUSE
10h45 - 11h15
Zenon Pylyshyn (Rutgers University, USA, psychologie cognitive)
What's in the Mind? A Computational Approach to a Ancient Question.
11h15 - 11h30 Discussion
11h30 - 12h00
Stevan Harnad (Princeton University, psychologie cognitive)
Modeles, mobiles et mentalite
12h00 - 12h15 Discussion

MEAL

14h00 - 14h30
Michel Imbert (Universite Paul Sabatier, Toulouse, neurosciences)
De l'etude du cerveau a la comprehension de l'esprit
14h30 - 14h45 Discussion
14h45 - 15h15
Guy Tiberghien (Univers Pierre Mendes-France,Grenoble,psychologie
cognitive)
Connexionnisme: stade supreme du behaviorisme ?
15h15 - 15h30 Discussion
15h30 - 15h45 PAUSE
15h45 - 16h15
Jacques Demongeot  (Universite Joseph Fourier, Grenoble, neurosciences)
Memoire d'evocation dans les reseaux de neurones
16h15 - 16h30 Discussion
16h30 - 17h00
Bennet Murdock (Universite de Toronto, psychologie cognitive)
THE ROLE OF FORMAL MODELS IN MEMORY RESEARCH
17h00 - 17h15 Discussion
17h15 - 17h45
Robert Proulx (Universite du Quebec a Montreal, neuro-psychologie)
Plausibilite biologique de certains systemes de categorisation adaptative a
base de reseaux de neurones
17h45 - 18h00 Discussion

TUESDAY, December 1
Session 2 : Epistemology, Philosophy of Mind and Cognition

9h - 9h30
Elisabeth Pacherie (Universite de Provence, CNRS & CREA, Paris)
Domaines cognitifs et modularite
9h30 - 9h45 Discussion
9h-45 - 10h15
Pierre Livet (Universite de Provence & CREA, Paris, philosophie)
Categorisation et connexionnisme
10h15 - 10h30 Discussion
10h30 - 10h45 PAUSE
Normand Lacharite (Universite du Quebec a Montreal, epistemologie)
10h45 - 11h15
Conflits de modeles en theorie de la representation
11h15 - 11h30 Discussion
11h30 - 12h00
Peter Gardenfors (Lund University, Suede, philosophie)
Language and the Evolution of Mind
12h15 - 12h15 Discussion

MEAL

14h00 - 14h30
Andy Clark (Washington University, philosophie)
Wild Cognition: Putting Representation in its Place
14h30 - 14h45 Discussion
14h45 - 15h15
Kevin Mulligan (Universite de Geneve, Suisse, philosophie)
Constance perceptuelle et contenu spatial
15h15 - 15h30 Discussion
15h30 - 15h45 PAUSE
15h45 - 16h15
Ronald De Sousa (Universite de Toronto, epistemologie)
La rationalite: un concept normatif ou descriptif ?
16h15 - 16h30 Discussion
16h30 - 17h00
Daniel Laurier (Universite de Montreal, philosophie)
Rationalite et naturalisme
17h00 - 17h15 Discussion
17h15 - 17h45
Joelle Proust (CNRS & CREA, Paris, philosophie)
Un modele naturaliste de l'intentionnalite
17h45 - 18h00 Discussion

FRIDAY, December 2
Session 3: Modelisation IA, Traitement du langage, et semantique cognitive

Paul Jorion (Maison des Sciences de l'Homme, Paris, psychologie cognitive)
9h - 9h30
Modelisation du reseau mnesique : une utilisation minimaliste de l'IA
9h30 - 9h45 Discussion
9h-45 - 10h15
Bernard Amy (Universite Joseph Fourier, Grenoble, connexionnisme)
La place des reseaux neuronaux dans l'IA
10h15 - 10h30 Discussion
10h30 - 10h45 PAUSE 
10h45 - 11h15
Paul Bourgine (CEMAGREF, Paris-Antony, IA-modelisation)
Co-evolution et emergence du soi
11h15 - 11h30 Discussion
11h30 - 12h00
Paul Pietroski (Universite McGill, Canada, philosophie)
What can linguistics teach us about belief
12h00 - 12h15 Discussion
MEAL
14h00 - 14h30
Le paradigme hermeneutique et la mediation semiotique
Francois Rastier (Institut National de la Langue Francaise, CNRS,
                  linguistique computationnelle)
14h30 - 14h45 Discussion
14h45 - 15h15
L'impact des perspectives cognitives dans le traitement de l'information
Jean-Guy Meunier (Universite du Quebec a Montreal, semiotique)
15h15 - 15h30 Discussion
15h30 - 15h45 PAUSE
15h45 -  16h15
Guy Denhiere (Universite Paris VIII, psychologie cognitive)
Isabelle Tapiero (Universite Lyon II, psychologie cognitive)
La signification comme structure emergente : de l'acces au lexique a la
comprehension de textes
16h15 - 16h30 Discussion
16h30 - 17h00
Paul Freedman (Centre de Recherche en Informatique de Montreal, IA)
La vision artificielle: le traitement intelligent de documents
17h00 - 17h15 Discussion
17h15 - 17h45
Denis Vernant (Universite Pierre Mendes-France, Grenoble, philosophie)
L'intelligence de la machine et sa capacite dialogique
17h45 - 18h00 Discussion

18h00: END OF COLLOQUIUM

------------------------------------------------------------------------
-ADMISSION FEES-
(Includes:access to the conference room, meals and the colloquium
documents)

Individuals-------------------------------------------------1500FF
Student (join proof of eligibility with registration)------- 500FF

----------------------------------------------------------------------------
--

                           REGISTRATION BULLETIN
(The Cognitive Sciences:From computational models to philosophy of mind)

Name:___________________________________________________________________
             
Status:_____________________________________                              
Institution/Company_________________________                               

Complete Address_________________________________________________________
Fax:________________________                                            
Phone :______________________                                     
@mail number__________________________________

Enclosed : 	 Check	 or money order of (_____________________FF)
(Make check or money order payable to CENTRE JACQUES CARTIER)

-Send information on possibilities of housing in Lyon(______)
_Send me the colloquium brochure (_____)

-November 30 meal __
-December 1, meal __
-December 2, meal __

RETURN TO: CENTRE JACQUES CARTIER, 86 rue Pasteur, 69365 Lyon Cedex 07,  
                                    France. Phone:(33) 78 69 72 21


From dhw at santafe.edu  Sat Aug 13 01:19:30 1994
From: dhw at santafe.edu (dhw@santafe.edu)
Date: Fri, 12 Aug 94 23:19:30 MDT
Subject: Cross-validation theory
Message-ID: <9408130519.AA08172@chimayo.santafe.edu>



Mark Plutowski recently said on connectionist:


>>>
A discussion on the Machine Learning List prompted the question

"Have theoretical conditions been established under which
cross-validation is justified?"

The answer is "Yes."
>>>

As Mark goes on to point out, there are decades
of work on cross-validation from a likelihood-driven
sampling theory perspective. Indeed, Mark's thesis is a
major addition to that literature.

Mark then correctly notes that this literature doesn't *directly* 
apply to the discusson on the ML list, since that discussion involves
off-training set rather than iid error.

It should be noted that there is another important distinction between 
the framework Mark uses and the implicit framework in the ML list 
discussion; the latter has been concerned w/ zero-one loss, whereas 
Mark's work concentrates on quadratic loss. The no-free-lunch results 
being discussed in the ML list change form drastically if one uses 
quadratic rather than zero-one loss. That should not be too surprising, 
given, for example, the results in Michael Perrone's work involving
quadratic loss.

It's also worth noting that even in the regime of iid error and quadratic
loss, there is still much to be understood. For example, in the 
average-data scenario of sampling theory statistics that Mark uses, 
asymptotic properties are better understood than finite data properties. 
And in the Bayesian this-data framework, very little is known for any 
data-size regime (though some work by Dawid on this subject comes to mind).




David Wolpert

From mdg at magi.ncsl.nist.gov  Tue Aug 16 11:58:28 1994
From: mdg at magi.ncsl.nist.gov (Mike Garris x2928)
Date: Tue, 16 Aug 94 11:58:28 EDT
Subject: Announcement: Public Domain OCR
Message-ID: <9408161558.AA02660@magi.ncsl.nist.gov>

                   ANNOUNCEMENT - PUBLIC DOMAIN OCR

             NIST FORM-BASED HANDPRINT RECOGNITION SYSTEM
             ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

              Michael D. Garris (mdg at magi.ncsl.nist.gov)
     James L. Blue, Gerald T. Candela, Darrin L. Dimmick, Jon Geist,
       Patrick J. Grother, Stanley A. Janet, and Charles L. Wilson

            National Institute of Standards and Technology,
                        Building 225, Room A216
                      Gaithersburg, Maryland 20899
            Phone: (301)975-2928          FAX: (301)840-1357


The National Institute of Standards and Technology (NIST) has developed a
standard reference form-based handprint recognition system for evaluating
optical character recognition (OCR). NIST is making this recognition system
freely available to the general public on an ISO-9660 format CD-ROM. The
recognition system processes the Handwriting Sample Forms distributed with
NIST Special Database 1 and NIST Special Database 3. The system reads
handprinted fields containing digits, lower case letters, upper case letters,
and reads a text paragraph containing the Preamble to the U.S. Constitution.

This is a source code distribution written primarily in C and is organized
into 11 libraries. There are approximately 19,000 lines of code supporting
more than 550 subroutines. Source code is provided for form registration,
form removal, field isolation, field segmentation, character normalization,
feature extraction, character classification, and dictionary-based post-
processing. A host of data structures and low-level utilities are also
provided. These utilities include the application of CCITT Group 4 decompres-
sion, IHead file manipulation, spatial histograms, Least-Squares fitting,
spatial zooming, connected components, Karhunen Loeve (KL) feature extraction,
optimized Probabilistic Neural Network classification, multiple-key sorting,
Levenstein distance dynamic string alignment, and dictionary-based post-
processing. Two supporting programs are provided that compute eigenvectors
and KL feature vectors for training classifiers. Unlike the recognition
system (which is written entirely in C), these two programs contain FORTRAN
subroutines. To support these programs, a training set of 168,365 segmented
and labeled character images is provided. About 1000 writers contributed to
this training set.

The NIST standard reference recognition system is designed to run on UNIX
workstations and has been successfully compiled and tested on a Digital
Equipment Corporation (DEC) Alpha, Hewlett Packard (HP) Model 712/80, IBM
RS6000, Silicon Graphics Incorporated (SGI) Indigo 2, SGI Onyx, SGI Challenge,
Sun Microsystems (Sun) IPC, Sun SPARCstation 2, Sun 4/470, and a Sun SPARC-
station 10.** Scripts for installation and compilation on these architectures
are provided with this distribution.

A CD-ROM distribution of this standard reference system can be obtained free
of charge by sending a letter of request to Michael D. Garris at the address
above. The letter, preferably on company letterhead, should identify the
requesting organization or individuals. This system or any portion of this
system may be used without restrictions. However, redistribution of this
standard reference recognition system is strongly discouraged as any
subsequent corrections or updates will be sent to registered recipients only.
This software was produced by NIST, an agency of the U.S. government, and by
statute is not subject to copyright in the United States. Recipients of this
software assume all responsibilities associated with its operation,
modification, and maintenance.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
** Specific hardware and software products identified were used in order to
   adequately support the development of this technology. In no case does
   such identification imply recommendation or endorsement by the National
   Institute of Standards and Technology, nor does it imply that the
   equipment identified is necessarily the best available for the purpose.

From terry at salk.edu  Tue Aug 16 20:12:13 1994
From: terry at salk.edu (Terry Sejnowski)
Date: Tue, 16 Aug 94 17:12:13 PDT
Subject: Neural Computation 6:5
Message-ID: <9408170012.AA07668@salk.edu>

NEURAL COMPUTATION
September 1994 Volume 6 Number 5

Articles:

A Bayesian Analysis of Self-Organizing Maps
Stephen P. Luttrell

Network Amplification of Local Fluctuations Causes High Spike Rate
Variability, Fractal Firing Patterns and Oscillatory Local Field Potentials
Marius Usher, Martin Stemmler, Christof Koch and Zeev Olami

Note:

Statistical Analysis of an Autoassociative Memory Network
A. M. N. Fu

Letters:

Loading Deep Networks is Hard
Jiri Sima

Measuring the VC-dimension of a Learning Machine
Vladimir Vapnik, Esther Levin and Yann Le Cun

Neural Nets with Superlinear VC-Dimension
Wolfgang Maass

A Novel Design Method for Multilayer Feedforward Neural Networks
Jihong Lee

An Internal Mechanism for Detecting Parasite Attractors in a Hopfield Network
Jean-Dominique Gascuel, Bahram Moobed and Michel Weinfeld

On Langevin Updating in Multilayer Perceptrons
Thorsteinn Rognvaldsson

Probabilistic Winner-Take-All Learning Algorithm for Radial-Basis-Function
Neural Classifiers
Hossam Osman and Moustafa M. Fahmy

Realization of the "Weak Rod" by a Double Layer Parallel Network
T. Matsumoto and K. Kondo

Learning in Neural Networks with Material synapses
Daniel J. Amit and Stefano Fusi

Model Based on Extracellular Potassium for Spontaneous Synchronous Activity
in Developing Retinas
Pierre-Yves Burgi and Norberto M. Grzywacz

Bayesian Modeling and Classification of Neural Signals
Michael S. Lewicki

-----
 
SUBSCRIPTIONS - 1994 - VOLUME 6 - BIMONTHLY (6 issues)
 
______ $40     Student and Retired
______ $65     Individual
______ $166    Institution
 
Add $22 for postage and handling outside USA (+7% GST for Canada).
 
(Back issues from Volumes 1-5 are regularly available for $28 each
to institutions and $14 each for individuals
Add $5 for postage per issue outside USA (+7% GST for Canada)
 
MIT Press Journals, 55 Hayward Street, Cambridge, MA 02142.
Tel: (617) 253-2889  FAX: (617) 258-6779  e-mail: hiscox at mitvma.mit.edu
 
-----

From ucganlb at ucl.ac.uk  Wed Aug 17 05:01:47 1994
From: ucganlb at ucl.ac.uk (Dr Neil Burgess - Anatomy UCL London)
Date: Wed, 17 Aug 94 10:01:47 +0100
Subject: pre-print in neuroprose: hippocampus - spatial models
Message-ID: <93386.9408170901@link-1.ts.bcc.ac.uk>


ftp://archive.cis.ohio-state.edu/pub/neuroprose/burgess.hbtnn.ps.Z

The above/below file has been put on neuroprose for anonymous ftp, www
or whatever, contact n.burgess at ucl.ac.uk with any retrieval problems.
All the best,
Neil

		HIPPOCAMPUS - SPATIAL MODELS
	Neil Burgess, Michael Recce & John O'Keefe
	Dept. of Anatomy, University College London,
		   London WC1E 6BT, U.K.
		e-mail: n.burgess at ucl.ac.uk

This is a brief review of models of the hippocampus, focussing on 
spatial aspects of cell-firing and hippocampal function. 
To appear in `The Handbook of Brain Theory and Neural Networks'
(M. A. Arbib Ed.), Bradford Books/ MIT Press, 1995, and restricted in 
length and number of citations accordingly.

8 pages, 0.46 Mbytes uncompressed, 
hard-copies availible in extreme circumstances only.


From uzimmer at informatik.uni-kl.de  Wed Aug 17 12:14:01 1994
From: uzimmer at informatik.uni-kl.de (Uwe R. Zimmer, AG vP)
Date: Wed, 17 Aug 94 17:14:01 +0100
Subject: Papers available (world modelling, mobile robots)
Message-ID: <940817.171401.723@ag_vp_file_server.informatik.uni-kl.de>

A couple of actual papers about:

--------------------------------------------------------------
---     Learning, Robotics, Visual Search, Navigation,     ---
---        Topologic Maps & Robust Mobile Robots           ---
---                    Neural Networks                     ---
--------------------------------------------------------------

are now available via FTP:

------------------------------------------------------------------------
--- Comparing World-Modelling Strategies for Autonomous Mobile Robots
------------------------------------------------------------------------

---   File name is : Zimmer.Comparison.ps.Z   ---

         IWK `94, Ilmenau, Germany, September 27 - 30, 1994

   Comparing World-Modelling Strategies for Autonomous Mobile Robots

                Uwe R. Zimmer & Ewald von Puttkamer

The focus of this paper is on strategies for adapting a couple of
internal representations to the actual environment of a mobile robot.

From svc at demos.lanl.gov  Tue Aug 16 15:22:10 1994
From: svc at demos.lanl.gov (Stephen Coggeshall)
Date: Tue, 16 Aug 94 13:22:10 MDT
Subject: graduate student positions
Message-ID: <9408161922.AA13573@demos.lanl.gov>

At Los Alamos National Laboratory in New Mexico we have a small research
effort using adaptive computational models for a variety of projects.  At
this time there may be the possibility of a few graduate student positions
available specifically for pattern recognition/data base mining with
application to financial problems. We are interested in highly motivated,
self-directed students with strong backgrounds in programming, math, neural
net applications.  Interested parties can contact Steve (svc at lanl.gov).
Please describe briefly your past work and current interests, as well as
your availability.

From R.Gaizauskas at dcs.shef.ac.uk  Fri Aug 19 17:12:12 1994
From: R.Gaizauskas at dcs.shef.ac.uk (Robert John Gaizauskas)
Date: Fri, 19 Aug 94 17:12:12 BST
Subject: AISB-95 WORKSHOP/TUTORIAL CALL
Message-ID: <9408191612.AA03843@dcs.shef.ac.uk>


        -------------------------------------------------
        AISB-95: CALL FOR WORKSHOP AND TUTORIAL PROPOSALS
        -------------------------------------------------

             Call for Workshop Proposals: AISB-95

                 University of Sheffield,
                    Sheffield, England

                    April 3 -- 4, 1995

                   Society for the Study of
      Artificial Intelligence and Simulation of Behaviour (SSAISB)


The AISB Committee invites proposals for workshops to be held in conjunction
with the Tenth Biennial Conference on AI and Cognitive Science (AISB-95).
While the main conference will run for three days from Wednesday, April 5 to
Friday, April 7, the workshops will be held on the two days preceding the main
event: Monday, April 3 and Tuesday, April 4.

The main conference has the theme "Hybrid Problems, Hybrid Solutions" (see the
main conference call) and while proposals for workshops related to that theme
would be particularly welcome, proposals are invited for workshops relating
to any aspect of Artificial Intelligence or the Simulation of Behaviour.

Proposals, from an individual or a pair of organisers, for workshops between
0.5 and 2 days long will be considered. Workshops will probably address topics
which are at the forefront of research, but not yet sufficiently developed to
warrant a full-scale conference.

 Submission:
 ----------
 A workshop proposal should contain the following information:

 1. Workshop Title

 2. A detailed outline of the workshop.
    This should include the necessary background and the potential
    target audience for the workshop and a justified estimate of the
    number of possible attendees. Please also state the length and
    preferred date(s) of the workshop. Specify any equipment requirements,
    indicating whether the organisers would be expected to meet them.

 3. A brief resume of the organiser(s).
    This should include: background in the research area,
    references to published work in the topic area and relevant
    experience, such as previous organisation or chairing of
    workshops.

 4. Administrative information.
    This should include: name, mailing address, phone number, fax, and
    email address if available.  In the case of multiple organisers,
    information for each organiser should be provided, but one organiser
    should be identified as the principal contact.

 5. A draft Call for Participation.
    This should serve the dual purposes of informing and attracting
    potential participants.


 The organisers of accepted workshops are responsible for issuing a call for
 participation, reviewing requests to participate and scheduling the workshop
 activities within the constraints set by the Workshop Organiser. They are
 also responsible for submitting a collated set of papers for their workshop
 to the Workshop Organiser.

 Dates:
 ------
 Intentions to organise a workshop should be made known to the
 Workshop Organiser as soon as possible.

 Proposals must be received by October 18th 1994.
 Decisions about topics and speakers will be made in early November.
 Collated sets of papers to be received by March 15th 1995.

 Proposals should be sent to:

    Dr. Robert Gaizauskas
    Department of Computer Science
    University of Sheffield
    211 Portobello Street
    Regent Court
    Sheffield S1 4DP
    U.K.

    email:  robertg at dcs.shef.ac.uk
    phone:  +44 (0)742 825572
    fax:    +44 (0)742 780972

Electronic submission (plain ascii text) is highly preferred, but
hard copy submission is also accepted, in which case 5 copies should
be submitted. Proposals should not exceed 2 sides of A4 (i.e. 120
lines of text approx.).

---------------------------------------------------------------------

             Call for Tutorial Proposals: AISB-95

                 University of Sheffield,
                    Sheffield, England

                    April 3 -- 4, 1995

                   Society for the Study of
      Artificial Intelligence and Simulation of Behaviour (SSAISB)


The AISB Committee invites proposals for Tutorials to be held in conjunction
with the Tenth Biennial Conference on AI and Cognitive Science (AISB-95).
While the main conference will run for three days from Wednesday, April 5 to
Friday, April 7, the tutorials will be held on the two days preceding the main
event: Monday, April 3 and Tuesday, April 4.

Proposals for full and half day tutorials, from an individual or pair of
presenters, will be considered. They may be offered both on standard topics
and on new and more advanced aspects of Artificial Intelligence or Simulation
of Behaviour. Anyone interested in presenting a tutorial should submit a
proposal to the Workshop Organiser Dr Robert Gaizauskas (addresses below).

 Submission:
 ----------
 A tutorial proposal should contain the following
 information:

 1. Tutorial Title

 2. A brief description of the tutorial, suitable for inclusion in
    a brochure.

 3. A detailed outline of the tutorial.
    This should include the necessary background and the potential
    target audience for the tutorial and a justified estimate of the
    number of possible attendees. Please also state the length and
    preferred date(s) of the tutorial. Specify any equipment requirements,
    indicating whether the organisers would be expected to meet them.

 4. A brief resume of the presenter(s).
    This should include: background in the tutorial area,
    references to published work in the topic area and relevant
    experience. Published work should, ideally, include a published
    tutorial-level article on the subject. Relevant experience
    is teaching experience, including previous conference tutorials
    or short courses presented.

 5. Administrative information.
    This should include: name, mailing address, phone number, fax, and
    email address if available.  In the case of multiple presenters,
    information for each presenter should be provided, but one presenter
    should be identified as the principal contact.

 The presenter(s) of accepted tutorials must submit a set of
 tutorial notes (which may include relevant tutorial-level
 publications) to the Workshop Organisers by March 15th 1995.

 Dates:
 ------
 Intentions to organise a tutorial should be made known to the
 the Workshop Organiser as soon as possible.

 Proposals must be received by October 18th 1994.
 Decisions about tutorial topics and speakers will be made in early November.
 Tutorial notes must be received by March 15th 1995.

 Proposals should be sent to:

    Dr. Robert Gaizauskas
    Department of Computer Science
    University of Sheffield
    211 Portobello Street
    Regent Court
    Sheffield S1 4DP
    U.K.

    email:  robertg at dcs.shef.ac.uk
    phone:  +44 (0)742 825572
    fax:    +44 (0)742 780972

Electronic submission (plain ascii text) is highly preferred, but
hard copy submission is also accepted, in which case 5 copies should
be submitted. Proposals should not exceed 2 sides of A4 (i.e. 120
lines of text approx.).

From bishopc at helios.aston.ac.uk  Fri Aug 19 09:33:57 1994
From: bishopc at helios.aston.ac.uk (bishopc)
Date: Fri, 19 Aug 1994 13:33:57 +0000
Subject: Research Associate - Software Support
Message-ID: <9192.9408191233@sun.aston.ac.uk>


-------------------------------------------------------------------


                     Aston University 

               Neural Computing Research Group


      RESEARCH ASSOCIATE - SYSTEMS AND SOFTWARE SUPPORT
      -------------------------------------------------


Applications are invited for a position as a Research Associate 
within the Neural Computing Research Group both to provide support for
the Group's local network of Sun workstations and to undertake 
software development and research in support of projects within
the Group. 

The Neural Computing Research Group currently comprises three
professors, two lecturers, three postdoctoral research fellows and 
ten postgraduate research students. In addition, two further 
Lecturerships have recently been advertised. Current research activity
focusses on principled approaches to neural computing, and spans 
a broad spectrum from theoretical foundations to industrial and
commercial applications. 

The ideal candidate will have significant experience of the UNIX
operating system and system maintenance, experience of software 
engineering in C++, and an understanding of neural networks.
The responsibilities of the successful candidate will be as 
follows:

(1) To provide system support for the Group's LAN of Sun UNIX
workstations and associated peripherals, ensurimg that an
efficient working environment is maintained.

(2) To support the development, testing and documentation of
the NetLib C++ library of neural network software.

(3) To assist with numerical experiments in support of research
projects, including industrial contracts. This aspect of the work 
may provide opportunities for joint publication in academic journals.

(4) To provide such other software support as may be required, 
such as the maintenance of LaTeX, provision of WWW pages for the
Group, etc.

Salaries will be 13,941 UK pounds or above, depending on
the experience and qualifications of the successful applicant.

If you wish to apply for this position, please send a CV, together 
with the names and addresses of 3 referees, to:

    Professor Chris Bishop
    Neural Computing Research Group
    Aston University
    Birmingham B4 7ET, U.K.
    Tel: 021 359 3611 ext. 4270
    Fax: 021 333 6215
    e-mail: c.m.bishop at aston.ac.uk

    closing date: 9 September 1994



From harnad at Princeton.EDU  Fri Aug 19 22:51:41 1994
From: harnad at Princeton.EDU (Stevan Harnad)
Date: Fri, 19 Aug 94 22:51:41 EDT
Subject: Brain Rhythms & Cognition: PSYCOLOQUY Call for Commentary
Message-ID: <9408200251.AA06967@clarity.Princeton.EDU>

  Pulvermueller et al: BRAIN RHYTHMS, CELL ASSEMBLIES AND COGNITION

The target article whose abstract appear below has just been just been
published in PSYCOLOQUY, a refereed electronic journal of Peer
Commentary sponsored by the American Psychological Association.
Formal commentaries are now invited. The full text can be easily and
instantly retrieved by a variety of simple means, described below.
Instructions for Commentators appear after the retrieval Instructions.

  TARGET ARTICLE AUTHOR'S RATIONALE FOR SOLICITING COMMENTARY:

  Fast periodic brain responses have been investigated in various
  mammals, humans included. Although most neuroscientists agree on the
  importance of these processes, it is not at all clear what role they
  play in cortical and subcortical processing. Are they simply a
  byproduct of perceptual processes, or do they play a role in what can
  be called higher or cognitive processing in the brain? We tried to
  answer this question by performing experiments in which spectral
  responses to meaningful words and physically similar but meaningless
  pseudowords were recorded from the human cortex. The result,
  differential 30-Hz responses to these stimuli, is interpreted in the
  framework of a Hebbian cell assembly theory. We hope that both the
  results and the brain-theoretic approach will stimulate a fruitful
  multidisciplinary discussion.

-----------------------------------------------------------------------
psycoloquy.94.5.48.brain-rhythms.1.pulvermueller  Friday 19 August 1994
ISSN 1055-0143   (30 paragraphs, 10 figs, 9 notes, 61 refs, 1203 lines)
PSYCOLOQUY is sponsored by the American Psychological Association (APA)
                Copyright 1994 Friedemann Pulvermueller et al.

        BRAIN RHYTHMS, CELL ASSEMBLIES AND COGNITION:
        EVIDENCE FROM THE PROCESSING OF WORDS AND PSEUDOWORDS

                Friedemann Pulvermueller (1)
                Hubert Preissl (1)
                Carsten Eulitz (2)
                Christo Pantev (2)
                Werner Lutzenberger (1)
                Thomas Elbert (2)
                Niels Birbaumer (1, 3)

    (1) Institut fuer Medizinische Psychologie und
        Verhaltensneurobiologie, Universitaet Tuebingen,
        Gartenstrasse 29, 72074 Tuebingen, Germany
        PUMUE at mailserv.zdv.uni-tuebingen.de

    (2) Institut fuer Experimentelle Audiologie, 
        Universitaet Muenster, Kardinal von Galen-Ring 10, 
        48149 Muenster, Germany

    (3) Universita degli Studi, Padova, Italy


    ABSTRACT: In modern brain theory, cortical cell assemblies are
    assumed to form the basis of higher brain functions such as form
    and word processing. When gestures or words are produced and
    perceived repeatedly by the infant, cell assemblies develop which
    represent these building blocks of cognitive processing. This leads
    to an obvious prediction: cell assembly activation ("ignition")
    should take place upon presentation of items relevant for cognition
    (e.g., words, such as "moon"), whereas no ignition should occur
    with meaningless items (e.g., pseudowords, such as "noom"). Cell
    assembly activity may be reflected by high-frequency brain
    responses, such as synchronous oscillations or rhythmic
    spatiotemporal activity patterns in which large numbers of neurons
    participate. In recent MEG and EEG experiments, differential
    gamma-band responses of the human brain were observed upon
    presentation of words and pseudowords. These findings are
    consistent with the view that fast coherent and rhythmic activation
    of large neuronal assemblies takes place with word but not
    pseudowords.

    KEYWORDS: brain theory, cell assembly, cognition, event related
    potentials (ERP), electroencephalograph (EEG), gamma band, Hebb,
    language, lexical processing, magnetoencephalography (MEG),
    psychophysiology, periodicity, power spectral analysis, synchrony


The article is retrievable from the following sites:

ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1994.volume.5
http://info.cern.ch/hypertext/DataSources/bySubject/Psychology/Psycoloquy.html
http://www.princeton.edu/~harnad/
gopher://gopher.cic.net/11/e-serials/alphabetic/p/psycoloquy
gopher://gopher.lib.virginia.edu/11/alpha/psyc
gopher://wachau.ai.univie.ac.at/11/archives/Psycoloquy

The filename is:
psycoloquy.94.5.48.brain-rhythms.1.pulvermueller

   To retrieve a file by ftp from a Unix/Internet site, type either:
ftp princeton.edu
   or
ftp 128.112.128.1
   When you are asked for your login, type:
anonymous
   Enter password as per instructions (your password is your own userid:
   yourlogin at yourhost.whatever.whatever - be sure to include the "@")
   and then change directories with:
cd pub/harnad/BBS
   To show the available files, type:
ls
   Next, retrieve the file you want with (for example):
get psycoloquy.94.5.48.brain-rhythms.1.pulvermueller
or
mget *pulv*
   When you have the file(s) you want, type:
quit
---------------------------------------------------------------- 
The file is also retrievable through archie, gopher, and World-Wide Web
using the URLs (Universal Resource Locators):

http://info.cern.ch/hypertext/DataSources/bySubject/Psychology/Psycoloquy.html
ftp://princeton.edu/pub/harnad/Psycoloquy
or
gopher://gopher.princeton.edu/1ftp%3aprinceton.edu%40/pub/harnad/BBS/
----------------------------------------------------------------
   Certain non-Unix/Internet sites have a facility you can use that is
   equivalent to the above. Sometimes the procedure for connecting to
   princeton.edu will be a two step process such as:
ftp
   followed at the prompt by:
open princeton.edu
  or
open 128.112.128.1
   In case of doubt or difficulty, consult your system manager.
------------------------------------------------------------------
Where the above procedures are not available (e.g. from Bitnet or other
networks), there are two fileservers --  ftpmail at decwrl.dec.com and
bitftp at pucc.bitnet -- that will do the transfer for you. Send either one the
one line message:
help
for instructions (which will be similar to the above, but will be in
the form of a series of lines in an email message that ftpmail or
bitftp will then execute for you).

--------------------------------------------------------------------------
        INSTRUCTIONS FOR PSYCOLOQUY AUTHORS AND COMMENTATORS

PSYCOLOQUY is a refereed electronic journal (ISSN 1055-0143) sponsored
on an experimental basis by the American Psychological Association
and currently estimated to reach a readership of 40,000. PSYCOLOQUY
publishes brief reports of new ideas and findings on which the author
wishes to solicit rapid peer feedback, international and
interdisciplinary ("Scholarly Skywriting"), in all areas of psychology
and its related fields (biobehavioral science, cognitive science,
neuroscience, social science, etc.). All contributions are refereed.

Target article length should normally not exceed 500 lines [c. 4500 words].
Commentaries and responses should not exceed 200 lines [c. 1800 words].

All target articles, commentaries and responses must have (1) a short
abstract (up to 100 words for target articles, shorter for commentaries
and responses), (2) an indexable title, (3) the authors' full name(s)
and institutional address(es).

In addition, for target articles only: (4) 6-8 indexable keywords,
(5) a separate statement of the authors' rationale for soliciting
commentary (e.g., why would commentary be useful and of interest to the
field? what kind of commentary do you expect to elicit?) and
(6) a list of potential commentators (with their email addresses).

All paragraphs should be numbered in articles, commentaries and
responses (see format of already published articles in the PSYCOLOQUY
archive; line length should be < 80 characters, no hyphenation).

It is strongly recommended that all figures be designed so as to be
screen-readable ascii. If this is not possible, the provisional
solution is the less desirable hybrid one of submitting them as
postscript files (or in some other universally available format) to be
printed out locally by readers to supplement the screen-readable text
of the article.

PSYCOLOQUY also publishes multiple reviews of books in any of the above
fields; these should normally be the same length as commentaries, but
longer reviews will be considered as well. Book authors should submit a
500-line self-contained Precis of their book, in the format of a target
article; if accepted, this will be published in PSYCOLOQUY together
with a formal Call for Reviews (of the book, not the Precis). The
author's publisher must agree in advance to furnish review copies to the
reviewers selected.

Authors of accepted manuscripts assign to PSYCOLOQUY the right to
publish and distribute their text electronically and to archive and
make it permanently retrievable electronically, but they retain the
copyright, and after it has appeared in PSYCOLOQUY authors may
republish their text in any way they wish -- electronic or print -- as
long as they clearly acknowledge PSYCOLOQUY as its original locus of
publication. However, except in very special cases, agreed upon in
advance, contributions that have already been published or are being
considered for publication elsewhere are not eligible to be considered
for publication in PSYCOLOQUY,

Please submit all material to psyc at pucc.bitnet or psyc at pucc.princeton.edu
Anonymous ftp archive is DIRECTORY pub/harnad/Psycoloquy HOST princeton.edu

From miku at sedal.su.oz.au  Tue Aug 23 03:02:45 1994
From: miku at sedal.su.oz.au (Michael Usher)
Date: Tue, 23 Aug 1994 17:02:45 +1000 (EST)
Subject: Sixth Australian Conference on Neural Networks
Message-ID: <199408230702.RAA26903@sedal.sedal.su.OZ.AU>

For those people putting the finishing touches to their submissions,
the Author's Style Guidelines are now available via the World Wide Web.
LaTeX style information can be found at URL:
	http://www.sedal.su.oz.au/acnn95

The conference programme and registration details will also be placed there,
when they become available.

Further queries about ACNN'95 should be directed to <acnn95 at ee.su.oz.au>.
Problems with the WWW server should be directed to myself.

Michael Usher
--
Michael Usher		Systems Administrator
miku at sedal.su.OZ.AU	Systems Engineering & Design Automation Lab (SEDAL)
Tel: +61 2 692 4135	Department of Electrical Engineering, Building J03
Fax: +61 2 660 1228	University of Sydney, NSW 2006, AUSTRALIA

From shultz at hebb.psych.mcgill.ca  Tue Aug 23 09:10:34 1994
From: shultz at hebb.psych.mcgill.ca (Tom Shultz)
Date: Tue, 23 Aug 94 09:10:34 EDT
Subject: No subject
Message-ID: <9408231310.AA19055@hebb.psych.mcgill.ca>

Subject: Paper available: A connectionist model of the learning of 
personal pronouns in English.
Date: 23 August '94

FTP-host: archive.cis.ohio-state.edu
FTP-file: pub/neuroprose/shultz.pronouns.ps.Z

-------------------------------------------------------------

The following paper has been placed in the Neuroprose archive at 
Ohio State University: 

A connectionist model of the learning of personal pronouns in 
English. (13 pages)

Thomas R. Shultz, David Buckingham, & Yuriko Oshima-Takane
Department of Psychology & McGill Cognitive Science Centre		
McGill University
Montreal, Quebec, Canada H3A 1B1		
shultz at psych.mcgill.ca

Abstract
Both experimental and observational psycholinguistic research have 
shown that children's acquisition of first and second person 
pronouns is affected by the opportunity to hear these pronouns used 
in speech not addressed to them. These effects were simulated with 
the cascade-correlation connectionist algorithm. The networks 
learned, in effect, to produce the pronouns "me" and "you" depending 
on identification of the speaker, addressee, and referent. Analysis of 
network performance and structure indicated that generalization to 
correct pronoun production was aided by listening to non-addressed 
speech and that persistent pronoun errors were created by listening 
to directly addressed speech. It was noted that explicit symbolic 
rule models would likely have difficulty simulating the pattern 
frequency effects common to the present simulations and to the 
natural language environment of the child.

The paper has been published in S. J. Hanson, T. Petsche, M. Kearns, & 
R. L. Rivest (Eds.) (1994). Computational learning theory and natural 
learning systems, Vol. 2: Intersection between theory and 
experiment (pp. 347-362). Cambridge, MA: MIT Press.  

Instructions for ftp retrieval of this paper are given below.  If you 
are unable to retrieve and print it and therefore wish to receive a 
hardcopy, please send e-mail to shultz at psych.mcgill.ca

Please do not reply directly to this message.

FTP INSTRUCTIONS:

unix> ftp archive.cis.ohio-state.edu (or 128.146.8.52)
    Name: anonymous
    Password: <your e-mail address>
    ftp> cd pub/neuroprose
    ftp> binary
    ftp> get shultz.pronouns.ps.Z 
    ftp> quit
unix> uncompress shultz.pronouns.ps.Z 

Thanks to Jordan Pollack for maintaining this archive. 

Tom Shultz

From tap at cs.toronto.edu  Tue Aug 23 16:18:48 1994
From: tap at cs.toronto.edu (Tony Plate)
Date: Tue, 23 Aug 1994 16:18:48 -0400
Subject: Thesis available for ftp
Message-ID: <94Aug23.161849edt.240@neuron.ai.toronto.edu>


Ftp-host: ftp.cs.utoronto.ca
Ftp-filename: /pub/tap/plate.thesis.2up.ps.Z
Ftp-filename: /pub/tap/plate.thesis.ps.Z

The following thesis is available for ftp.  There are two
versions: plate.thesis.ps.Z prints on 216 sheets of
paper, and plate.thesis.2up.ps.Z prints on 108 sheets.
The compressed files are around 750Kb each.

	Distributed Representations and
	Nested Compositional Structure

		    by

		Tony A. Plate*
	Department of Computer Science,
	   University of Toronto,
		   1994

A thesis submitted in conformity with the requirements for
the degree of Doctor of Philosophy in the Graduate Department
of Computer Science at the University of Toronto.


		 Abstract

Distributed representations are attractive for a number of
reasons.  They offer the possibility of representing
concepts in a continuous space, they degrade gracefully with
noise, and they can be processed in a parallel network of
simple processing elements.  However, the problem of
representing nested structure in distributed representations
has been for some time a prominent concern of both
proponents and critics of connectionism
\cite{fodor-pylyshyn-88,smolensky-90,hinton-90}.  The lack
of connectionist representations for complex structure has
held back progress in tackling higher-level cognitive tasks
such as language understanding and reasoning.

In this thesis I review connectionist representations and
propose a method for the distributed representation of
nested structure, which I call ``Holographic Reduced
Representations'' (HRRs).  HRRs provide an implementation of
Hinton's~\shortcite{hinton-90} ``reduced descriptions''.
HRRs use circular convolution to associate atomic items,
which are represented by vectors.  Arbitrary variable
bindings, short sequences of various lengths, and
predicates can be represented in a fixed-width vector.
These representations are items in their own right, and can
be used in constructing compositional structures.  The noisy
reconstructions extracted from convolution memories can be
cleaned up by using a separate associative memory that has
good reconstructive properties.

Circular convolution, which is the basic associative
operator for HRRs, can be built into a recurrent neural
network.  The network can store and produce sequences.  I
show that neural network learning techniques can be used
with circular convolution in order to learn representations
for items and sequences.

One of the attractions of connectionist representations of
compositional structures is the possibility of computing
without decomposing structures.  I show that it is possible
to use dot-product comparisons of HRRs for nested structures
to estimate the analogical similarity of the structures.
This demonstrates how the surface form of connectionist
representations can reflect underlying structural similarity
and alignment.


* New Address:

Tony Plate,
Department of Chemical Engineering
University of British Columbia
2216 Main Mall, Vancouver,
BC, Canada V6T 1Z4
tap at chml.ubc.ca

From d.rosen at ieee.org  Wed Aug 24 10:47:12 1994
From: d.rosen at ieee.org (d.rosen@ieee.org)
Date: Wed, 24 Aug 94 07:47:12 -0700
Subject: Software Engineer/Developer (N.Y.) [Neural Networks] Wanted
Message-ID: <9408241445.AA14964@solstice>

Full-time position for Software Engineer/Developer with recent Degree.
Possible alternative: Part-time Experienced Software Engineer/Developer.

Starting date: Immediate/September.

A federally-funded research team at New York Medical College is
applying neural networks and advanced probabilistic/statistical
methods to improve the accuracy with which the stage (of advancement)
of cancer cases can be evaluated -- an important factor in determining
treatment.

We seek a skilled developer to take primary responsibility for the
design and implementation of our neural network software, which will
be geared towards flexible experimental use in our fast-paced research
program, as well as a simple GUI prototype of clinical production
software.  The successful candidate will work with us (medical, neural
network, and statistical researchers) to plan the best path from our
current C code to a more carefully-designed, extensible, OO approach,
perhaps with partial rapid implementation in an interpreted language,
and evaluate the possible role of other available tools and libraries.
It is expected that eventually, much of the resulting software will be
freely distributed for use in many fields.

Candidates should have demonstrably outstanding skills in designing
object-oriented software, and in C++ (or both C and some OO language)
development under the Unix[/X11] environment.  Prefer knowledge of as
many of the following as possible: neural nets, statistics, numerical
/ scientific computation, portable GUI, prototyping language (Python,
Smalltalk, Perl, S-plus, ...), MS Windows, and Unix system
administration.

New York Medical College (NYMC) is located in the community of
Valhalla, NY, just half an hour north of New York City.  The position
would be on-site, though doing some portion of the work remotely could
perhaps be arranged.

NYMC is the third-largest private medical university in the United
States. It is an Equal Opportunity and Affirmative Action Institution.
Currently we do not believe we could justify hiring an individual who
is not already authorized to work in the U.S.

If you are interested and qualified, please e-mail your resume to me
as soon as possible (plain text preferred).

--
David Rosen, PhD   <d.rosen at ieee.org>

From iiscorp at netcom.com  Wed Aug 24 18:48:07 1994
From: iiscorp at netcom.com (IIS Corp)
Date: Wed, 24 Aug 94 15:48:07 PDT
Subject: Soft Computing Days in San Francisco
Message-ID: <CMM.0.90.4.777768487.iiscorp@netcom>



              Soft Computing Days in San Francisco


           Zadeh, Widrow, Koza, Ruspini, Stork, Whitley, 
                  Bezdek, Bonissone, and Berenji

                              On

              
                         Soft Computing:  
	 Fuzzy Logic, Neural Networks, and Genetic Algorithms

     
                      Three short courses


                       San Francisco, CA
                       October 24-28, 1994


Traditional (hard) computing methods do not provide sufficient
capabilities to develop and implement intelligent systems.  Soft
computing methods have proved to be important practical tools to build
and construct these systems.  

The following three courses, offered by Intelligent Inference Systems 
Corp., will focus on all major soft computing technologies: fuzzy logic,
neural networks, genetic algorithms, and genetic programming. These
courses may be taken either individually or in combination.

Course 1: Artificial Neural Networks (Oct. 24)
          Bernard Widrow and David Stork

Course 2: Genetic Algorithms and Genetic Programming (Oct. 25)
	  John Koza and Darrell Whitley

Course 3: Fuzzy Logic Inference (Oct. 26-28)
	  Lotfi Zadeh, Jim Bezdek, Enrique Ruspini, 
	  Piero Bonissone, and Hamid Berenji

For further details on course topics and registration information, send
an email to iiscorp at netcom.com or contact Intelligent Inference
Systems Corp., Phone (415) 988-9934, Fax: (415) 988-9935.  A detailed
brochure will be sent to you as soon as possible.


From ken at phy.ucsf.edu  Thu Aug 25 08:16:49 1994
From: ken at phy.ucsf.edu (ken@phy.ucsf.edu)
Date: Thu, 25 Aug 1994 05:16:49 -0700
Subject: postdoctoral position available in computational neuroscience
Message-ID: <9408251216.AA20136@coltrane.ucsf.EDU>

Christof Schreiner works on the physiology of auditory cortex.  His
lab has defined several of the auditory parameters that seem to be
mapped in auditory cortex.  He has funding for a postdoctoral position
for theoretical work aimed at understanding the mapping to auditory
cortex and its possible consequences for representation of complex
sounds.  His description follows:

	The objective is to develop network models of the mammalian auditory
	cortex on the basis of a broad data base of physiological observations
	from our laboratory. Based on the spectral-temporal filter properties
	of neurons and their spatial distribution in the cortex, consequences
	for the cortical representation of complex signals (such as animal
	vocalizations and speech) shall be evaluated. Previous experience in
	signal processing and self-organizing or other neural network classes
	is required.

Christof and I are both members of the Keck Center for Integrative
Neuroscience at UCSF, a group of 10 faculty working on systems
neuroscience including Michael Stryker, Michael Merzenich, Steve
Lisberger, Allison Doupe, Alan Basbaum, Roger Nicoll, Howard Fields,
and Henry Ralston.  The postdoc will work in the Keck Center, and
both I and Christof will be available to work closely with them.

DO NOT SEND YOUR APPLICATIONS TO ME!!

Send applications to:

Christof Schreiner
Dept. of Otolaryngology
Box 0732
UCSF
513 Parnassus
SF, CA 94143-0732

email: chris at phy.ucsf.edu

Please send a cv, and names, addresses and phone numbers of three
individuals who can provide references for you.  Copies of your
publications would also be helpful.

Ken Miller

From bishopc at helios.aston.ac.uk  Thu Aug 25 07:18:00 1994
From: bishopc at helios.aston.ac.uk (bishopc)
Date: Thu, 25 Aug 1994 11:18:00 +0000
Subject: NIPS Workshop - Inverse Problems
Message-ID: <4310.9408251018@sun.aston.ac.uk>


-------------------------------------------------------------------

                      NIPS*94 Workshop:


                     DOING IT BACKWARDS: 
                     -------------------
 
      NEURAL NETWORKS AND THE SOLUTION OF INVERSE PROBLEMS
      ----------------------------------------------------
 
                  Organizer: Chris M Bishop



                    CALL FOR PRESENTATIONS
                    ---------------------- 


Introduction:
------------

Many of the tasks for which neural networks are commonly used 
correspond to the solution of an `inverse' problem. Such tasks are 
characterized by the existence of a well-defined, deterministic 
`forward' problem which might, for instance, correspond to causality 
in a physical system. By contrast the inverse problem may be 
ill-posed, and may exhibit multiple solutions. 


Aims:
----

A wide range of different approaches have been developed to tackle
inverse problems, and one of the main goals of the workshop is to 
contrast the way in which they address the underlying technical issues, 
and to identify key areas for future research.


Format:
------

This will be a one-day workshop, and will involve short presentations,
with ample time allowed for discussions. 


Contributions
-------------

If you wish to make a contribution to this workshop, please e-mail a 
brief outline of your proposed presentation to c.m.bishop at aston.ac.uk 
by 5 September.


Chris Bishop


-------------------------------------------------------------------- 

  Professor Chris M Bishop           Tel. +44 (0)21 359 3611 x4270
  Neural Computing Research Group    Fax. +44 (0)21 333 6215
  Dept. of Computer Science          c.m.bishop at aston.ac.uk
  Aston University               
  Birmingham B4 7ET, UK

--------------------------------------------------------------------


From oby at TechFak.Uni-Bielefeld.DE  Thu Aug 25 13:13:10 1994
From: oby at TechFak.Uni-Bielefeld.DE (oby@TechFak.Uni-Bielefeld.DE)
Date: Thu, 25 Aug 94 19:13:10 +0200
Subject: No subject
Message-ID: <9408251713.AA02980@gaukler.TechFak.Uni-Bielefeld.DE>


POSTDOCTORAL POSITION

Postdoctoral position available beginning spring / summer 1995 in a project
combining anatomy, physiology and theoretical modelling to understand the
functional architecture of primate visual cortex. Funded by the Human
Frontiers program, the position is initially for two years and the
postdoctoral fellow will be expected to interact with members of the group
in Germany, England, and USA.


The candidate should have some experience with neural modeling and possess
appropriate math and and computer competency; the candidate should also
have relevant experimental skills (eg anatomy or physiological unit
recording) in mammalian visual system.


Applicants should send their CV, list of publications, letter describing
their interest in the position, and name,address and phone number of two
referees to either


Prof. Jennifer Lund, 

Institute of Ophthalmology, 

11-43 Bath St. , London EC1V 9EL,UK 

Phone (0)71-608-6864), E-mail smgxjsl at ucl.ac.uk; 


or 


Dr. Klaus Obermayer, 

Universitaet Bielefeld, Technische Fakultat,
Universitaetsstrasse 25, 33615 Bielefeld, Germany 

Phone (0)521-106-6058; Email oby at techfak.uni-bielefeld.de.


From ccg at melissa.lanl.gov  Thu Aug 25 18:23:12 1994
From: ccg at melissa.lanl.gov (Camilo Gomez)
Date: Thu, 25 Aug 94 16:23:12 MDT
Subject: positions in financial research
Message-ID: <9408252223.AA00343@melissa.lanl.gov.demosdiscs>

(PLEASE POST)

-----------------------------------------------------------------------------------------

               POSITIONS IN FINANCIAL RESEARCH
                             AT
               LOS ALAMOS NATIONAL LABORATORY
                CENTER FOR NON-LINEAR STUDIES
             (Postdoctoral and Graduate Student)

Positions  involving research and development  of  financial
models  are  available  at Los Alamos  National  Laboratory.
Depending  on  funding we will have a  number  of  positions
including:

     a)postdoctoral
     b)graduate student

The  successful  candidates will  be  expected  to  work  on
projects involving research and development in the  area  of
financial  derivatives. These will involve both fixed-income
and  equity  derivatives. Projects will focus  on  valuation
problems  for  a  number  of  these  financial  derivatives.
Exceptionally well qualified candidates with an interest  in
computational investigations of above mentioned  topics  and
expertise in one or more of the following or related  areas,
are encouraged to apply:

     a)finance
     b)financial derivatives
     c)statistical analysis
     d)time series analysis
     e)neural net/pattern recognition
     f)emergent behavior systems
     g)parallel computing
     h)programming skills (C and C++ languages)

Candidates may contact:


     M.F. Gomez
     CNLS, MS-B258
     Los Alamos National Laboratory
     Los Alamos, NM 87544
     frankie at cnls.lanl.gov

for application material and questions.

Please indicate in your initial inquiry that is for a
position in financial research and whether you are
interested in a student or post-doctoral position.

Los Alamos National Laboratory is an equal opportunity
employer.

-----------------------------------------------------------------------------------------

From shannon at cis.ohio-state.edu  Thu Aug 25 15:02:52 1994
From: shannon at cis.ohio-state.edu (shannon roy campbell)
Date: Thu, 25 Aug 1994 15:02:52 -0400 (EDT)
Subject: paper available on synchrony and desynchrony in oscillator an oscillator network
Message-ID: <199408251902.PAA27060@axon.cis.ohio-state.edu>


FTP-host: archive.cis.ohio-state.edu
FTP-filename : /pub/neuroprose/campbell.wc_oscillators.ps.Z

The file campbell.wc_oscillators.ps.Z (34 pages) is now 
available for copying from the Neuroprose repository. 
Contact shannon at cis.ohio-state.edu for retrieval problems.

          
           Synchronization and Desynchronization in a 
        Network of Locally Coupled Wilson-Cowan Oscillators

            by Shannon Campbell and DeLiang Wang~
               Dept. of Physics 
              ~Dept. of Computer and Information Science
               Ohio State University
               Columbus, Ohio 43210-1277

Abstract - 
   A network of Wilson-Cowan oscillators is constructed, and its
emergent properties of synchronization and desynchronization are 
investigated by both computer simulation and formal analysis. The 
network is a two-dimensional matrix where each oscillator is
coupled only to its neighbors. We show analytically that a chain of
locally coupled oscillators (the piece-wise linear approximation to 
the Wilson-Cowan oscillator) synchronizes, and present a technique 
to rapidly entrain finite numbers of oscillators. The coupling
strengths change on a fast time scale based on a Hebbian rule. A
global separator is introduced which receives input from and sends 
feedback to each oscillator in the matrix. The global separator is 
used to desynchronize different oscillator groups. Unlike many 
other models, the properties of this network emerge from local 
connections, that preserve spatial relationships among components 
and are critical for encoding Gestalt principles of feature 
grouping. The ability to synchronize and desynchronize oscillator
groups within this network offers a promising approach for pattern 
segmentation and figure/ground segregation based on oscillatory
correlation. 

FTP INSTRUCTIONS:

unix> ftp archive.cis.ohio-state.edu (or 128.146.8.52)
    Name: anonymous
    Password: <your e-mail address>
    ftp> cd pub/neuroprose
    ftp> binary
    ftp> get campbell.wc_oscillators.ps.Z 
    ftp> quit
unix> uncompress campbell.wc_oscillators.ps.Z 

From shannon at cis.ohio-state.edu  Fri Aug 26 10:25:48 1994
From: shannon at cis.ohio-state.edu (shannon roy campbell)
Date: Fri, 26 Aug 1994 10:25:48 -0400 (EDT)
Subject: correct reference for paper on synchrony and desynchrony...
Message-ID: <199408261425.KAA27616@axon.cis.ohio-state.edu>


        Sorry that I did not specify the proper reference to a previously
        announced paper in neuroprose ("Synchrony and desynchrony in
        a network of Wilson-Cowan oscillators"). The paper is a techical
        report from The Department of Computer and Information Science,
        The Ohio State University, numbered "OSU-CISRC-8/94-TR43".

        Thanks for your attention.
        -shannon campbell

From kreider at bechtel.Colorado.EDU  Sat Aug 27 16:21:38 1994
From: kreider at bechtel.Colorado.EDU (Jan Kreider)
Date: Sat, 27 Aug 1994 14:21:38 -0600
Subject: System identification competition announcement
Message-ID: <199408272021.OAA10216@bechtel.Colorado.EDU>


Announcement of 

***System Identification Competition***

Benchmark tests for estimation methods of thermal characteristics of 
buildings and building components.


*Objective

The objective of the benchmark is to set-up a comparison between 
alternative techniques and to clarify particular problems of system 
identification applied to the thermal performance of buildings. 


*Organisation

J. Bloem, U. Norlen, EC - Joint Research Centre, Ispra, Italy
H. Madsen, H. Melgaard, IMM TU of Denmark, Lyngby, Denmark
J. Kreider, JCEM, University of Colorado, U.S.


*Active period of the competition : July 1, 1994 - December 31, 1994


*Introduction

A wide range of system identification techniques is now being applied 
to the analysis problems involved with estimation of thermal properties 
of buildings and building components. Similar problems arise in most 
observational disciplines, including physics, biology, and economics. 
New commercially available software tools and special purpose computer 
programs promise to provide results that were unobtainable just a 
decade ago. Unfortunately, the realisation and evaluation of this 
promise has been hampered by the difficulty of making rigorous 
comparisons between competing techniques, particularly ones that come 
from different disciplines.

This competition has been organised to help clarify the conflicting 
claims among many researchers who use and analyse building energy data 
and to foster contact among these persons and their institutions. The 
intent is not necessarily only to declare winners but rather to set up 
a format in which rigorous evaluations of techniques can be made. 
Because there are natural measures of performance, a rank-ordering will 
be given. In all cases, however, the goal is to collect and analyse 
quantitative results in order to understand similarities and differences 
among the approaches.

At the close of the competition the performance of the techniques 
submitted will be compared.  Those with the best results will be asked 
to write a scientific paper and will be invited for a presentation of 
the paper. There will be no monetary prizes. A symposium at the JRC 
Ispra, Northern Italy, has been scheduled for the Autumn 1995 to 
explore the results of the competition in formal papers. The competition,
 the overall results and papers on selected methods will published by 
the organisers in a book.

Research on energy savings in buildings can be divided in three major 
areas:

1) building components, 2) test cells and unoccupied buildings in 
real climate and 3) occupied buildings. Three competitions are planned 
along this line of which the present competition concerned with building 
components will be the first one.

The present competition is concerned with wall components and no solar 
radiation involved. Five different cases are provided for estimation and
prediction. Four cases have been designed with wall components in order
to test parameter estimation methods. Prediction tests are also 
included. Some of the dependent variable values will be withheld from 
the data set in these cases. Contestants are free to submit results 
from any number of cases.

When the outcome of this first competition is positive a second 
competition is planned which concerns test cells and unoccupied 
buildings under real climate conditions (1995). A third competition 
concerns occupied buildings (1996).

If there is sufficient interest, a network server may be set up to 
operate as an on-line archive of interesting data sets, programs, and 
comparisons among algorithms in the future. 


***ACCESSING THE DATA

The competition does not require advanced registration; there are two 
ways to enter: 

1. by normal mail. Simply request the data by sending a letter. The 
data are available on diskettes (3.5-in size) in ASCII, IBM-PC format. 
(there is no charge for the data diskette). To receive the data, 
send the letter together with a self-addressed rigid envelope to :

Joint Research Centre
Institute of System Engineering and Informatics
J.J. BLOEM,     Building 45
I - 21020 Ispra (VA), ITALY

2. by E-mail. In that case just send an request for participation by 
E-mail to J. Kreider at the University of Colorado, Boulder, 
CO 80309-0428, U.S. at the following E-mail address: 

jkreider at vaxf.colorado.edu

Information how to obtain the necessary instructions and the required 
data series, using FTP, are forwarded to you by E-mail.
Instructions on submitting a return disk with the analysis of the cases 
will be included in a README file. The disk will also include an entry 
form that each participant will need to complete and submit along with 
the results.

***FOR MORE INFORMATION


Further questions about the competition should be directed to one of 
the following organisers:


Joint Research Centre
Institute for Systems Engineering 
and Informatics
J.J. BLOEM      Building 45
I - 21020 ISPRA (VA), Italy
tel:    +39 332 789842/789145
fax:    +39 332 789992
E-mail:  hans.bloem at cen.jrc.it


Joint Center for Energy Management
J. KREIDER
Campus Box 428
University of Colorado
Boulder, CO  80309-0428, U.S.
tel:    +303 492 3915
fax:    +303 492 7317
E-mail:  jkreider at vaxf.colorado.edu


End of Announcement SysId Competition.


From lautrup at connect.nbi.dk  Mon Aug 29 15:54:23 1994
From: lautrup at connect.nbi.dk (Benny Lautrup)
Date: Mon, 29 Aug 94 15:54:23 METDST
Subject: no subject (file transmission)
Message-ID: <mailman.700.1149540294.24850.connectionists@cs.cmu.edu>

Subject: Paper available: Extremely Ill-posed Learning
Date: August 29, 1994

FTP-host: connect.nbi.dk
FTP-file: neuroprose/hansen.ill-posed.ps.Z

----------------------------------------------

The following paper is now available:

Extremely Ill-posed Learning [14 pages]

L.K. Hansen, B. Lautrup, I. Law, N. Moerch, and J. Thomsen
CONNECT, The Niels Bohr Institute, University of Copenhagen, Denmark
 
Abstract:
Extremely ill-posed learning problems  are  common  in  image  and
spectral analysis. They are characterized  by  a  vast  number  of
highly correlated inputs, eg pixel  or  or  pin  values,  and  a
modest number of patterns, eg images or spectra. We show that it
is possible to  train  neural  networks  to  learn  such  patterns
without using an excessive number of weights, and we devise a test
to decide if new patterns should be included in the  training  set
or whether they fall within the  subspace  already  explored.  The
method is applied to the analysis of PET-images.

Please do not reply directly to this message.

-----------------------------------------------

FTP-instructions:

unix> ftp connect.nbi.dk (or 130.225.212.30)
ftp> Name: anonymous
ftp> Password: your e-mail address
ftp> cd neuroprose
ftp> binary
ftp> get hansen.ill-posed.ps.Z
ftp> quit
unix> uncompress hansen.ill-posed.ps.Z
        
-----------------------------------------------

Benny Lautrup,
Computational Neural Network Center (CONNECT)
Niels Bohr Institute
Blegdamsvej 17
2100 Copenhagen
Denmark

Telephone: +45-3532-5200
Direct:    +45-3532-5358
Fax:       +45-3142-1016
e-mail:    lautrup at connect.nbi.dk

From marshall at cs.unc.edu  Mon Aug 29 12:03:50 1994
From: marshall at cs.unc.edu (Jonathan A. Marshall)
Date: Mon, 29 Aug 1994 12:03:50 -0400
Subject: POSTDOC JOB (Neural net models of learning)
Message-ID: <199408291603.MAA27461@marshall.cs.unc.edu>

**********************************************************************

POSTDOCTORAL POSITION 

Available in September 1994.  Focus of research is neural network
models of learning, with a special emphasis on spatial tasks.

Qualified Ph.D. applicants should send a cover letter indicating
research experience, vitae, and names of three references to

Dr. Nestor Schmajuk
Department of Psychology
Duke University
Durham, NC 27706

e-mail: nestor at acpub.duke.edu

**********************************************************************

From ccg at melissa.lanl.gov  Tue Aug 30 11:01:01 1994
From: ccg at melissa.lanl.gov (Camilo Gomez)
Date: Tue, 30 Aug 94 09:01:01 MDT
Subject: positions in financial research (corrected e-mail)
Message-ID: <9408301501.AA02500@melissa.lanl.gov.demosdiscs>

(PLEASE POST)
-----------------------------------------------------------------------------------------

               POSITIONS IN FINANCIAL RESEARCH
                             AT
               LOS ALAMOS NATIONAL LABORATORY
                CENTER FOR NON-LINEAR STUDIES
             (Postdoctoral and Graduate Student)

Positions  involving research and development  of  financial
models  are  available  at Los Alamos  National  Laboratory.
Depending  on  funding we will have a  number  of  positions
including:

     a)postdoctoral
     b)graduate student

The  successful  candidates will  be  expected  to  work  on
projects involving research and development in the  area  of
financial  derivatives. These will involve both fixed-income
and  equity  derivatives. Projects will focus  on  valuation
problems  for  a  number  of  these  financial  derivatives.
Exceptionally well qualified candidates with an interest  in
computational investigations of above mentioned  topics  and
expertise in one or more of the following or related  areas,
are encouraged to apply:

     a)finance
     b)financial derivatives
     c)statistical analysis
     d)time series analysis
     e)neural net/pattern recognition
     f)emergent behavior systems
     g)parallel computing
     h)programming skills (C and C++ languages)

Candidates may contact:


     M.F. Gomez
     CNLS, MS-B258
     Los Alamos National Laboratory
     Los Alamos, NM 87544
     frankie at goshawk.lanl.gov

for application material and questions.

Please indicate in your initial inquiry that is for a
position in financial research and whether you are
interested in a student or post-doctoral position.

Los Alamos National Laboratory is an equal opportunity
employer.

-----------------------------------------------------------------------------------------

From prechelt at ira.uka.de  Tue Aug 30 15:32:35 1994
From: prechelt at ira.uka.de (Lutz Prechelt)
Date: Tue, 30 Aug 1994 21:32:35 +0200
Subject: Report on NN learning algorithm evaluation practices
Message-ID: <"irafs2.ira.198:30.07.94.19.32.14"@ira.uka.de>


A technical report titled

  A Study of Experimental Evaluations
  of Neural Network Learning Algorithms:
  Current Research Practice

is now available for anonymous ftp (binary mode!) from

  archive.cis.ohio-state.edu  /pub/neuroprose/prechelt.eval.ps.Z

The file is 70 Kb, the paper has 9 pages.

This is the abstract:

113 journal articles about neural network learning algorithms published in
1993 and 1994 are examined for the amount of experimental evaluation they
contain. Every third of them does employ not even a single realistic or real
learning problem. Only 6\% of all articles present results for more
than one problem using real world data.
Furthermore, one third of all articles does not present any
quantitative comparison with a previously known algorithm.
These results indicate that the quality of research in the
area of neural network learning algorithms needs improvement.
The publication standards should be raised and easily accessible
collections of example problems be built.

Here is a bibtex entry for the report:

@techreport{Prechelt94d,
   author    = {Lutz Prechelt},
   title     = {A Study of Experimental Evaluations of Neural Network
                Learning Algorithms: Current Research Practice},
   institution = {Fakult\"at f\"ur Informatik, Universit\"at Karlsruhe},
   year      = {1994},
   number    = {19/94},
   address   = {D-76128 Karlsruhe, Germany},
   month     = aug,
   note      = {anonymous ftp: /pub/papers/techreports/1994/1994-19.ps.Z
                on ftp.ira.uka.de},
}

Comments are welcome.

  Lutz
  
Lutz Prechelt   (email: prechelt at ira.uka.de)            | Whenever you 
Institut fuer Programmstrukturen und Datenorganisation  | complicate things,
Universitaet Karlsruhe;  76128 Karlsruhe;  Germany      | they get
(Voice: ++49/721/608-4068, FAX: ++49/721/694092)        | less simple.

From tgd at chert.CS.ORST.EDU  Tue Aug 30 19:07:33 1994
From: tgd at chert.CS.ORST.EDU (Tom Dietterich)
Date: Tue, 30 Aug 94 16:07:33 PDT
Subject: paper: Error Correcting Output Codes
Message-ID: <9408302307.AA13478@edison.CS.ORST.EDU>

The following paper is available at URL:
"ftp://ftp.cs.orst.edu/pub/tgd/papers/tr-ecoc.ps.gz" 

               Solving Multiclass Learning Problems via
                    Error-Correcting Output Codes

                Thomas G. Dietterich  tgd at cs.orst.edu
                   Department of Computer Science,
                          303 Dearborn Hall
                       Oregon State University
                       Corvallis, OR 97331 USA

                            Ghulum Bakiri
                    Department of Computer Science
                        University of Bahrain
                          Isa Town, Bahrain


    Multiclass learning problems involve finding a definition for an
    unknown function f(x) whose range is a discrete set containing k>2
    values (i.e., k ``classes'').  The definition is acquired by studying
    large collections of training examples of the form <x_i, f(x_i)>.
    Existing approaches to multiclass learning problems include (a) direct
    application of multiclass algorithms such as the decision-tree
    algorithms C4.5 and CART, (b) application of binary concept learning
    algorithms to learn individual binary functions for each of the $k$
    classes, and (c) application of binary concept learning algorithms
    with distributed output representations such as those employed by
    Sejnowski and Rosenberg in the NETtalk system.  This paper compares
    these three approaches to a new technique in which error-correcting
    codes are employed as a distributed output representation.  We show
    that these output representations improve the generalization
    performance of both C4.5 and backpropagation on a wide range of
    multiclass learning tasks.  We also demonstrate that this approach is
    robust with respect to changes in the size of the training sample, the
    assignment of distributed representations to particular classes, and
    the application of overfitting avoidance techniques such as
    decision-tree pruning.  Finally, we show that--like the other
    methods--the error-correcting code technique can provide reliable
    class probability estimates.  Taken together, these results
    demonstrate that error-correcting output codes provide a
    general-purpose method for improving the performance of inductive
    learning programs on multiclass problems.


Thomas G. Dietterich              Voice: 503-737-5559
Department of Computer Science    FAX:   503-737-3014
Dearborn Hall, 303                URL:   http://www.cs.orst.edu/~tgd/index.html
Oregon State University
Corvallis, OR 97331-3102

From georgiou at wiley.csusb.edu  Wed Aug 31 14:00:01 1994
From: georgiou at wiley.csusb.edu (georgiou@wiley.csusb.edu)
Date: Thu, 1 Sep 1994 02:00:01 +0800
Subject: JCIS: Last Call for Papers
Message-ID: <9409010900.AA04051@orion.csci.csusb.edu>

	       JOINT CONFERENCE ON INFORMATION SCIENCES

			 Last Call for Papers

(For submission of Neural Networks papers please see end of message for
address.  Deadline: September 10, 1994)


ORGANIZERS 

Honorary Chairs                   
Lotfi A. Zadeh & Azriel Rosenfeld 

Managing Chair of the Joint Conferences 
Paul P. Wang 
Dept. of Electrical Engineering 
Duke University 
Durham, NC 27708-0291 
Tel: (919) 660 5271, 660-5259 
Fax: (919) 660-5293, 684-4860 
e-mail: ppw at ee.duke.edu 

Advisorv Board 
Nick Baker 
Earl Dowell 
Erol Gelenbe 
Stephen Grossberg 
Kaoru Hirota 
Abe Kandel 
George Klir 
Teuvo Kohonen 
Tosiyasu L. Kunii 
Jin-Cherng Lin 
E. Mamdani 
Xuan-Zhong Ni 
C.V Ramamoorthy 
John E.R Staddon 
Masaki Togai 
Victor Van Beuren 
Max Woodbury 
Stephen S. Yau 
Lotfi Zadeh 
H. Zimmerman 

Keynote Speakers 
Lotfi A. Zadeh & Stephen Grossberg 

Plenarv Speakers 
Suguru Arimoto 
Dennis Bahler 
James Bowen 
Abe Kandel 
George Klir 
Phillipe Smets 
John R. Rice 
l.B. Turksen 
Benjamin Wah 
Stephen S. Yau

General Information 

The JOINT CONFERENCE ON INFORMATION SCIENCES consists of two international
conferences and one workshop. All interested attendees including
researchers, organizers, speakers, exhibitors, students and other
participants should register either in Plan A: 3rd International Conference
on Fuzzy Theory and Technology or Plan B: First International Conference on
Computer Theory & lnformatics and Workshop on Mobile Computing Svstems. Any
participants can attend all the keynote speeches, plenary sessions, all
parallel sessions and exhibits. The only difference is that all authors
registered in Plan A will participate in Lotfi A. Zadeh BestPaper
Competition. Plan B will have no best paper competition; at least for this
year. In addition, each plan will publish its own proceeding.

Tutorials 

Session A: Fuzzy Theory & Technology
(Sunday, November 13, 1994, 8:30 am - 12:30 pm)

1. George  Klir .................................Fuzzy Set and  Logic 
2. l. B. Turksen ................................Fuzzy Expert Systems 
3. Jack Aldridge .......................................Fuzzy Control 
4. Marcus Thint .......................Fuzzy Logic and NN Integration

Session B: Computers 
(Sunday, November 13, 1994, 1:30pm- 6:30pm)

1. Richard Palmer ....................................Neural Network 
2. Frank Y. Shih ................................Pattern Recognition 
3. Patrick Wang ......Intelligent Pattern Recognition & Applications 
4., 5. Ken W. White...............Success with Machine Vision I & II

Time Schedule & Venue

Tutorials ....................November 13, 1994 o 8:30 am - 6:30 pm 
Conferences .............November 14 - 16, 1994 o 8:30 am - 5:30 pm 
Venue........Pinehurst Resort & Country Club, Pinehurst, NC, U.S.A.

PARTICIPATION PLAN A:
3rd Annual Conference on Fuzzy Theory and Technology

PROGRAM COMMITTEE

Jack Aldridge 
Suguru Arimoto 
W Bandler 
P Bonnisone 
Bruno Bosacchi 
B Bouchon-Meunier 
J Buckley 
Dev Garg 
Rhett George 
George Georgiou 
I R Goodman 
Siegfried Gottwald 
Silvia Guiasu 
M M. Gupta 
Ralph Horvath 
D L Hung 
Timothy Jacobs 
Y.K Jani 
Joaquim Jorge 
Paul Kainen 
S.C. Kak 
Abe Kandel 
P Klement 
L J Kohout 
Vladik Kreinovlch 
N Kuroki 
Reza Langari
Harry Hua Li 
Don Meyers 
C K Mitchell 
John Mordeson 
Akira Nakamura 
Kyung Whan Oh 
Maria Orlowska 
Robert M. Pap 
Arthur Rarner 
Elie Sanchez 
B Scwhott 
Shouyoe Shao 
Sujeet Shenoi 
Frank Shih 
H Allison Smith 
L M Sztandera 
Alade ToKuta 
R Tong 
I Turksen 
Guo Jun Wang 
Tom Whalen 
Edward K Wong 
T Yamakawa 

The conference will consist of both plenary sessions and contributory
sessions, focusing on topics of critical interest and the direction of
future research.  Example topics include, but are not limited to the
following:

TOPICS: 3rd ANNUAL CONFERENCE ON FUZZY THEORY AND TECHNOLoGY

o Fuzzy Mathematics
o Basic Principles and Foundations of Fuzzy Logic
o Qualitative and Approximate-Reasoning Modeling
o Hardware Implementations of Fuzzy Logic Algorithms
o Design . Analysis, and Synthesis of Fuzzy Logic Controllers
o Learning and Acquisition of Approximate Models
o Fuzzy Expert Systems          
o Neural Network Architectures
o Artificially Intelligent Neural Networks
o Artificial Life
o Associative Memory
o Computational Intelligence
o Cognitive Science
o Fuzzy Neural Systems
o Relations between Fuzzy Logic and Neural Networks
o Theory of Evolutionary Computation 
o Efficiency/robustness comparisons with other direct search algorithms
o Parallelcomputer:applications
o Integration of Fuzzy Logic and Evolutionary Computing
o Comparisons between different variants of evolutionary algorithms 
o Evolutionary Computation for neural networks
o Fuzzy logic in Evolutionary algorithms
o Neurocognition
o Neurodynamics
o Optirnization
o Pattern Recognition
o Learning and Memory
o Machine Learning Applications
o Implementations (electronic, Optical, biochips)
o Intelligent Control

APPLICATIONS OF THE TOPICS:

o Hybrid Systems
o Image Processing
o Image Understanding
o Pattern Recognition
o Robotics and Automation
o Intelligent Vehicle and Highway Systems 
o Virtual Reality
o Tactile Sensors 
o Machine Vision 
o Motion Analysis
o Nuro biology
o Sensation and Perception
o Sensorimotor Systems
o Speech, Hearing and Language 
o Signal Processing
o Time Series Analysis
o Prediction
o System ldentification  
o System Control
o Intelligent Information Systems
o Case-Based Reasoning 
o Decision Analysis 
o Databases and Information Retrieval
o Dynamic Systems Modeling & Diagnosis
o Electric & Nuclear Power Systems

PARTICIPATION PLAN B:
1st Annual Computer Theory and Informatics Conference
& Workshop on Mobile Computing Systems

PROGRAM COMMITTEE:
FIRST ANNUAL, COMPUTER & INFORMATICS

Rafi Ahmed 
R Alonso 
Suguru Arimoto 
B. Badrinath 
C.R. Baker 
Martin Boogaard 
O Bukhres 
P Chrysantis 
Eliseo Clementini 
E.M. Ehlers 
Ahmed Elmagarmid 
K Ferentinos 
Godfrey 
Mohamed Gouda 
Albert G. Greensberg 
S Helal 
T Imielinski 
Subhash C Kak 
Abe Kandel 
Teuvo Kohomen 
Timo Koski 
Devendra Kumar 
Tosiyasu L Kunii 
Shahram Latifi 
Lin-shan Lee 
Mark Levene 
Jason Lin 
M i Lu 
Yanni  Manolopoulos 
Jorge Muruzabal 
Sham B Navathe 
Sigeru Ornatu 
C V Ramamoorthy 
Hari N. Reddy 
John R Rice 
Abdellah Salhi 
Frank S.Shih 
Harpreet Singh 
Stanley Y W. Su 
Abdullah Uz Tansel 
Kishor Trivedi 
Millist Vincent 
Benjamin Wah 
Z.A Wahab 
Jun Wang 
Patrick Wang 
Edward  K. Wong 
Lotfi A Zadeh

TOPICS: 1st ANNUAL COMPUTER THEORY  & INFORMATICS CONFERENCE

The conference will consist of both plenary sessions and contributory
sessions, focusing on topics of critical interest and the direction of
future research.  Example topics include, but are not limited to the
following:

o Computational Theory: Coding theory, automata, information theory, modern
algebraic theory, measure theory, probability and statistics, and numerical
methods

o Design of Algorithms: algorithmic complexity, theory of algorithms,
design, analyis and evaluation of efficient algorithms for engineering
applications (such as computer-aided design, computational fluid dynamics,
computer graphics, and virtual reality), combinatorics, scheduling theory,
discrete optimization, data compression, and approximation theory

o Software Design: Formal languages, theory and design of optimizing
compilers (especially those for parallel and supercomputers),
object-oriented programming, database theory and data organization,
software design methodology, program verification, and software
reliability.

o Computer systems and architectures: parallel and distributed computing
systems, high speed computer networks, theory and data orgarization,
software design methodology, program verification, and software
reliability.

o Evaluation methods and tools: Performance evaluation methods,
visualization tools, and simulation theory and methodology.

TOPICS: WORKSHOP ON MOBILE COMPUTING SYSTEMS

The workshop will focus on system support for mobile information and data
access, to recognize the role of mobile computer systems in today's
business and scientific communities. The workshop will be organized to
gather leading researchers and practitioners in the field. We shall focus
on issues related but not limited to: 

o Architectures for Mobile Computing Systems 
o Mobile ComputingTechnology 
o Wireless Communications 
o User lnterfaces in Palmtop Computing 
o Databases for Nomadic Computing 
o Transaction Models and Management 
o System Cornplexity, Integrity and Security
o Legal/Social/Health Issues  
o Operating System Support 
o Battery Management 
o Nomadic Applications
o Handheld Multimedia
o Personai Communication Networks

PROGRAM COMMITTEE: WORKSHOP ON MOBILE COMPUTING SYSTEMS 

R Alonso (Technology) 
B Badrinath (Query Processing) 
O Bukhres (Database Systems) 
P Chrysantis (Transactions) 
S Helal (Applications) 
T Imielinski (New Directions)

Note:
All Attendees must choose either Participation Plan A or Participation Plan B

PUBLICATIONS

The conference publishes two Proceedings on Summary; One entitled "Third
International Conference on Fuzzy Theory and Technology" and the other
entitled "First International Conference on Computer Theory and
Informatics, and First Workshop on Mobile Computing Systems." Both
proceedings will be made available on November 13,1994.

A summary shall not exceed 4 pages of 10-point font, doublecolumn,
single-spaced text, (1 page minimum) with figures and tables
included. Any summary exceeding 4 pages will be charged $100 per
additional page. Three copies of the summary are required by September
10, 1994. It is very important to mark "plan A" or "plan B" on your
manuscript. The conference will make the choice for you if you forget
to do so. Final version of the full length paper must be submitted by
November 14, 1994. Four (4) copies of the full length paper shall be
prepared according to the "Information for Authors" appearing at the
back cover of Information Sciences, an International Journal (Elsevier
Publishing Co.). A full paper shall not exceed 20 pages including
figures and tables. All full papers will be reviewed by experts in
their respective fields. Revised papers will be due on April 15, 1995.
Accepted papers will appear in the hard-covered proceeding (book) with
uniform typesetting to be published by a publisher (there will be two
books published this year, one for each plan) or Information Sciences
Journal (INS journal now has three publications: Informatics and
Computer Science, Intelligent Systems, Applications). All fully
registered conference attendees will receive a copy of proceeding
(summary) on November 14, 1994; a free one-year subscription (paid by
this conference) of Information Sciences Journal -
Applications. Lastly, the right to purchase either or both
hard-covered, deluxe, professional books at 1/2 price. The Title of
the books are "Advances in Fuzzy Theory & Technology, Volume lll",
"Advances in Computer Science and Informatics, Volume 1."

Lotfi A. Zadeh "Best Paper Award" FT&T 1994

All technical papers subrnitted to FT & T, 1994 are automatically qualified
as candidates for this award. The prize for this award is $2,500 plus hotel
accommodations (traveling expenses excluded) at FT & T, 1995. The date for
announcement of th best paper is May 30,1995. Oral presentation in person
at FT & T, 1994, is required and an acceptance speech at FT & T, 1995 is
also required. The evaluation committee for FT & T, 1994 consists of the
following 10 members:

Jack Aldridge 
B. Bouchon-Meunier 
George Klir l.R. Goodman 
John Mordeson 
Sujeet Shenoi 
H. Chris Tseng 
Frank Y. Shih 
Akira Nakamura 
Edward K. Wong 
I. B. Turksen.

The selection of the top ten best papers will be decided by conference
attendees and session chairs jointly.

EXHIBITIONS

JCIS '94 follows on last year's highly successful exhibits by some major
publishers, most publishers, lead by Elsevier, will return. Intelligent
Machines, Inc. will demonstrate their highly successful new software
"O'inca"-a FL-NN, Fuzzy-Neuro Design Framework. Dr. Ken W. White of Ithaca,
N.Y. will demonstrate his visual-sense systems. Virtus - a virtual reality
company based in Cary, North Carolina has committed to participate.
Negotiations are also underway with UNCVR research laboratory for its
participation. This conference intends to develop "virtual reality" as one
of the themes to benefit all attendees. Interested potential contributors
should contact Dr. Paul P. Wang or Dr. Rhett T. George.

Interested vendors should contact: Rhett George, E.E. Dept., Duke
University Telephone: (919) 660-5242 Fax: (919) 660-5293 rtg at ee.duke.edu

TRAVEL ARRANGEMENTS

The Travel Center of Durham, Inc. has been designated the official travel
provider. Special domestic fares have been arranged and The Travel Center
is prepared to book all flight travel. Domestic United States and Canada:1
-800-334-1 085 International FAX: (919) 687-0903

HOTEL RESERVATIONS Pinehurst Resort & Country Club Pinehurst, North
Carolina, U.S.A.

This is the conference site and lodging. Group Reservation Request designed
specifically for our conference. Very Special discount rates have been
agreed upon. Daily Rates for Hotel:

Single Occupancy - $122.00 
Double Occupancy - $91.00 per person 
Daily Rates for Manor Inn: Single Occupancy - $108.00 
Double Occupancy - $79.00/person 
(Rates are per person, per night and include accommodations, breakast and
dinner daily.)

Pinehurst Resort encompasses an elegant historic hotel (registered with
Historic Hotels of America") with the best in accommcations, gourmet dining
and modern meeting facilities. Our AAA Four Diamond and Mobil Four Star
resort offers a wide range of activities including seven championship golf
courses, tennis, waters sports, croquet and sport shooting.

Please contact: 
JACKIE HAYTER 
Associate Director of Sales 
Pinehurst Resort and Country Club 
Carolina Vista 
P.O. Box 4000 
Pinehurst, NC 28374-4000 
(919) 295-1339  - (919) 295-8484 
1-800-659-G0LF 

SPONSORS

Machine Intelligence and Fuzzy Logic Laboratory o Department of Electrical
Engineering, Duke University o Elsevier Science Inc. New York, N.Y.

PARTICIPANTS

IFSA, International Fuzzy Systems Association o Institute Of Information
science, Academia Sinica.

JCIS '94 REGISTRATION FEES & INFO.

                           Up to 9/15/94   After 9/15/94 
Full Registration            $275.00          $395.00 
Student Registration          $85.00          $160.00 
Tutorial w/ Conf_ Reg.       $150.00          $200.00 
Tutorial w/o Conf. Reg.      $300.00          $500.00 
Exhibit Booth Fee            $400.00          $500.00
One Day Fee 
(no pre-reg. discount)       $165.00 Full     $80.00 Student

Above fees applicable to both Plan A & Plan B

FULL CONFERENCE REGISTRATION: Includes admission to all sessions, exhibit
area, coffee, tea and soda A copy of conference proceedings (summany) at
conference and one year subscnption of Information Sciences - Applications,
An International Journal, published by Elsevier Publishing Co. In addition,
the right to purchase the hard-cover deluxe books at 1/2 price. Banquets
(November 14 & 15, 1994) are included through Hotel Registration. Tutorials
are not included

STUDENT CONFERENCE REGISTRATION. For full-time students only. A letter from
your department is required. You must present a current student I D with
picture. A copy of Conference Proceedings (Summary) is included. Admission
to all sessions, exhibit area, coffee, tea and soda. The nght to purchase
the hard-cver deluxe books at 1/2 price Free Subscription of INS Journal -
Applications is not included.

TUTORIALS REGISTRATION: Any person can register for the Tutorials A copy of
lecture notes for the oaurse registered is included. Coffee, tea and soda
is included The summary and tree subscription to the INS journal is,
however, not included. The right to purchase hard-cover deluxe books is
included.

VARIOUS CONFERENCE CONTACTS:

LOCAL INFORMATlON                                 
Rhett T. George Dept. of Electrical Engineering Box 90291, Duke University,
Durham, NC 27708-0291 e-mail: rtg at ee.duke.edu Tel. (919) 660-5228

TUTORIAL & CONFERENCE INFORMATION 
Paul P. Wang                          Kitahiro Kaneda 
e-mail: ppw at ee.duke.edu               e-mail: hiro at ee.duke.edu    
Tel. (919) 660-5271, 660-5259         Tel. (919) 660-5233 

Jerry C.Y. Tyan 
e-mail: ctyan at ee.duke.edu 
Tel. (919) 660-5233             
----------------------------------------------------------------

Neural Networks papers: Send summaries to

George M. Georgiou
Computer Science Department                  TEL: (909) 880-5332
California State University	             FAX: (909) 880-7004
5500 University Pkwy
San Bernardino, CA 92407, USA
georgiou at silicon.csci.csusb.edu

Deadline: September 10, 1994

TeX/Latex or postcript by email is fine.
--------------------------------------------------------------------


From todd at phy.ucsf.edu  Mon Aug  1 12:30:59 1994
From: todd at phy.ucsf.edu (todd@phy.ucsf.edu)
Date: Mon, 1 Aug 1994 09:30:59 -0700
Subject: New paper: Ffwd Hebbian Learning w/ Nonlinear Outputs
Message-ID: <9408011630.AA04378@dizzy.ucsf.EDU>



FTP-host: archive.cis.ohio-state.edu
FTP-filename: /pub/neuroprose/troyer.ffwd_hebb.ps.Z

The following paper has been submitted to Neural Networks and is
available from the Ohio State neuroprose archive. 

TITLE:   Feedforward Hebbian Learning with Nonlinear Output Units:
         A Lyapunov Approach

AUTHOR:  Todd Troyer <todd at phy.ucsf.edu>

         W.M. Keck Center for Integrative Neuroscience
            and Department of Pysiology
         513 Parnassus Ave. Box 0444
         University of California, San Francisco
         San Francisco, CA.  94143

ABSTRACT:

A Lyapunov function is constructed for the unsupervised learning
equations of a large class of two layer networks.  Units in the output
layer are recurrently connected by fixed symmetric weights; only the
feedforward connections between layers undergo learning.  In contrast
to much of the previous work on self-organization of this type, the
output units have nonlinear transfer functions.  The Lyapunov function
is similar in form to that derived by Cohen-Grossberg and Hopfield.
Two theorems are proved regarding the location of stable equilibria in
the limit of high gain transfer functions.  The analysis is applied to
the soft competitive learning networks of Amari and Takeuchi.


Retrieve this paper by anonymous ftp from:
archive.cis.ohio-state.edu (128.146.8.52)
    in the /pub/neuroprose    directory

The name of the paper in this archive is:
    troyer.ffwd_hebb.ps.Z          [15 pages]

No hard copies available.




From giles at research.nj.nec.com  Mon Aug  1 13:12:11 1994
From: giles at research.nj.nec.com (Lee Giles)
Date: Mon, 1 Aug 94 13:12:11 EDT
Subject: Reprint:SYNAPTIC NOISE IN DYNAMICALLY-DRIVEN RECURRENT NEURAL NETWORKS
Message-ID: <9408011712.AA13061@fuzzy>



****************************************************************************************

Reprint:SYNAPTIC NOISE IN DYNAMICALLY-DRIVEN RECURRENT NEURAL NETWORKS:
        CONVERGENCE AND GENERALIZATION

The following reprint is available via the University of Maryland Department of Computer 
Science Technical Report archive:

_________________________________________________________________________________________

          "Synaptic Noise in Dynamically-driven Recurrent Neural Networks: 
                       Convergence and Generalization"

      UNIVERSITY OF MARYLAND TECHNICAL REPORT UMIACS-TR-94-89 AND CS-TR-3322
               
                     Kam Jim(a), C.L. Giles(a,b), B.G. Horne(a)

                     {kamjim,giles,horne}@research.nj.nec.com

        (a) NEC Research Institute,4 Independence Way, Princeton, NJ 08540
  (b) Institute for Advanced Computer Studies, U. of Maryland, College Park, MD 20742  


There has been much interest in applying noise to feedforward neural networks in order 
to observe their effect on network performance. We extend these results by introducing 
and analyzing various methods of injecting synaptic noise into dynamically-driven 
recurrent networks during training. By analyzing and comparing the effects of these 
noise models on the error function, we found that applying a controlled amount of noise 
during training can improve convergence time and generalization performance. In addition, 
we analyze the effects of various noise parameters (additive vs. multiplicative, 
cumulative vs. non-cumulative, per time step vs. per sequence) and predict that best 
overall performance can be achieved by injecting additive noise at each time step. Noise 
contributes a second-order gradient term to the error function which can be viewed as an 
anticipatory agent to aid convergence. This term appears to find promising regions of 
weight space in the beginning stages of training when the training error is large and 
should improve convergence on error surfaces with local minima.Synaptic noise also 
enhances the error function by favoring internal representations where state nodes are 
operating in the saturated regions of the sigmoid discriminant function, thus improving 
generalization to longer sequences. We substantiate these predictions by performing 
extensive simulations on learning the dual parity grammar from grammatical strings 
encoded as temporal sequences with a second-order fully recurrent neural network.

--------------------------------------------------------------------------------------

                          FTP INSTRUCTIONS

                unix> ftp cs.umd.edu (128.8.128.8)
                Name: anonymous
                Password: (your_userid at your_site)
                ftp> cd pub/papers/TRs
                ftp> binary
                ftp> get 3322.ps.Z
                ftp> quit
                unix> uncompress 3322.ps.Z

--------------------------------------------------------------------------------------


--                                 
C. Lee Giles / NEC Research Institute / 4 Independence Way
Princeton, NJ 08540 / 609-951-2642 / Fax 2482
==




From inbs at bingsuns.cc.binghamton.edu  Mon Aug  1 13:48:58 1994
From: inbs at bingsuns.cc.binghamton.edu (INBS-conference)
Date: Mon, 1 Aug 94 13:48:58 EDT
Subject: call for papers
Message-ID: <9408011748.AA23972@bingsuns.cc.binghamton.edu >


                 C A L L        for      P A P E R S

          First International IEEE Symposium on  


       INTELLIGENCE in NEURAL AND BIOLOGICAL SYSTEMS


          May 29-31,1995, Washington DC Area, USA
        
           Sponsored by: IEEE Computer Society; 


In Cooperating with: AAAS Society; AAAI Society; SMC Society


AIMS and SCOPE of the Symposium

The Intelligence in Artificial Neural Networks and the Computational 
evolution of the Biological Systems are two very important and very 
active research areas, which offer and promise many practical 
applications to scientists and other professionals in industry 
and goverment as well. In response to this demand, the INBS 
Symposium offers a theoretical and a practical medium for the 
evolutionary and the intelligence processes in both artificial and 
biological systems and the interaction between these fields.

Some Topics 

 EVOLUTIONARY COMPUTING (DNA sequence processing, Genome Processes,  
  DNA, Topologies, Synthesis of DNA, Formal Linguistic of DNA, 
  Structure/Function Correlation,   Computational Genetics)
 BIOTECHNOLOGY AND APPLICATIONS (Mapping the Genome, Human Genome,      
  Molecular Computing, Limitations of Biological Models)
 GENETIC ALGORITHMS (Clustering, Optimization, Searching, Programming, etc)
 LANGUAGES UNDERSTANDING (Natural Languages, NL Translation, 
  Text Abstraction, Computational Linguistics)
 LEARNING AND PERCEPTION (Supervised, Unsupervised, Hybrid, 
  Understanding, Planning, Interpretation)
 NEUROSCIENCE (Adaptive Control Models, Fuzzy & Probabilistic Models, 
  Hybrid Models, Dynamic Neural and Neurocomputing Models, 
  Self-Organized Models)

SOME KEYNOTE DISTIGUISHED SPEAKERS

A.Apostolico, Europe;  S.Arikawa, Japan; S.Hameroff, USA


PROGRAM COMMITTEE                      Local Arrangement/Publicity Committee
N.Bourbakis, BU-SUNY,USA, Chair        Cynthia Shapiro
R.Brause, UG,Germany                   Sukarno Mertoguno
K.DeJong, GMU,USA                      James Gattiker
J.Shavlik, UWM,USA                     Ali Moghaddamzadeh
C.Koutsougeras, TU,USA
H.Kitano, JAPAN                                
M.Perlin, CMU,USA                      Publication Registration Committee
H.Pattee, BU,USA                       D.I. Kavraki
D.Schaffer, Philips Lab,USA                     
D.Searls, UPA,USA                               
J.Collado-Vides, UNAM
T.Yokomori, UEC,Japan
S.Rasmussen, Los Alamos,NL
G.Paun, Roumania
A.Restivo, U.Palermo,Italy
M.Chrochmore, U.Paris, France
D.Perrin, U.Paris, France
R.Reynolds, WSU,USA
M.Conrad, WSU,USA
M.Kameyama, TU,Japan
J.Nikolis, UP,Greece
T.Head, BU, USA, Vice Chair
C.Benham, MS,USA
R.VanBuskirk, BU-SUNY,USA
E.Dietrich, BU-SUNY,USA
S.Koslow, NIH,USA
M.Huerta, NIH,USA
B.Punch, MSU,USA



INFORMATION FOR AUTHORS
Authors are requested to submit four copies (in English) of 
their typed complete manuscript (25 pages max), or an 
extended summary (5-10 pages max) by Nov. 31,1994, to 
N. G. Bourbakis, Binghamton University,T.J.Watson School,
AAAI Lab, NY 13902,Tel: 607-777-2165, 607-771-4033; 
Fax:607-777-4822, E-mail : Bourbaki at BingSuns.CC.Binghamton.edu, 
or INBS at Bingsuns.CC.Binghamton.edu.
Each manuscript submitted to INBS must indicate the most relevant 
areas and include the complete address of at least one of the 
authors. Notification of acceptance, Jan. 31,1995. Final
copies of the papers accepted by INBS due March 21 ,1995.


From tesauro at watson.ibm.com  Mon Aug  1 13:50:48 1994
From: tesauro at watson.ibm.com (tesauro@watson.ibm.com)
Date: Mon, 1 Aug 94 13:50:48 EDT
Subject: NIPS*94 Registration
Message-ID: <mailman.694.1149591311.29955.connectionists@cs.cmu.edu>

Registration for NIPS*94 is now open.  A registration brochure
is available on-line via the NIPS*94 Mosaic homepage
(http://www.cs.cmu.edu:8001/afs/cs/project/cnbc/nips/NIPS.html),
or by anonymous FTP:
FTP site: mines.colorado.edu (138.67.1.3)
FTP file: /pub/nips94/nips94-registration-brochure.ps

The brochure contains the registration form and describes
the program highlights, including the list of invited speakers
and tutorial speakers.

People without access to FTP or Mosaic may request a copy of
the brochure by e-mail to "nips94 at mines.colorado.edu" or by
physical mail to:

        NIPS*94 Registration
        Dept. of Mathematical and Computer Sciences
        Colorado School of Mines
        Golden, CO  80401  USA


-- Gerry Tesauro
   NIPS*94 General Chair

From plaut at cmu.edu  Tue Aug  2 10:14:54 1994
From: plaut at cmu.edu (David Plaut)
Date: Tue, 02 Aug 1994 10:14:54 -0400
Subject: TR: Understanding Normal and Impaired Word Reading
Message-ID: <7478.775836894@crab.psy.cmu.edu>

FTP-host:	hydra.psy.cmu.edu [128.2.248.152]
FTP-file:	pub/pdp.cns/pdp.cns.94.5.ps.Z

[78 pages; 353Kb compressed; 924Kb uncompressed]

For those who do not have FTP access, physical copies can be requested from
Barbara Dorney <bd1q+ at andrew.cmu.edu>.

============================================================================

		Understanding Normal and Impaired Word Reading:
	       Computational Principles in Quasi-Regular Domains

	     David C. Plaut                    James L. McClelland
	Carnegie Mellon University          Carnegie Mellon University

	    Mark S. Seidenberg                 Karalyn E. Patterson
      University of Southern California     MRC Applied Psychology Unit

			 Technical Report PDP.CNS.94.5
				   July 1994

We develop a connectionist approach to processing in quasi-regular domains, as
exemplified by English word reading.  A consideration of the shortcomings of a
previous implementation (Seidenberg & McClelland, 1989, Psych. Rev.)  in
reading nonwords leads to the development of orthographic and phonological
representations that capture better the relevant structure among the written
and spoken forms of words.  In a number of simulation experiments, networks
using the new representations learn to read both regular and exception words,
including low-frequency exception words, and yet are still able to read
pronounceable nonwords as well as skilled readers.  A mathematical analysis of
the effects of word frequency and spelling-sound consistency in a related but
simpler system serves to clarify the close relationship of these factors in
influencing naming latencies.  These insights are verified in subsequent
simulations, including an attractor network that reproduces the naming latency
data directly in its time to settle on a response.  Further analyses of the
network's ability to reproduce data on impaired reading in surface dyslexia
support a view of the reading system that incorporates a graded
division-of-labor between semantic and phonological processes.  Such a view is
consistent with the more general Seidenberg and McClelland framework and has
some similarities with---but also important differences from---the standard
dual-route account.

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
David Plaut			plaut at cmu.edu	      "Doubt is not a pleasant
Department of Psychology	412/268-5145	       condition, but certainty
Carnegie Mellon University	412/268-5060 (FAX)     is an absurd one."
Pittsburgh, PA  15213-3890	345H Baker Hall			--Voltaire
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

From john at dcs.rhbnc.ac.uk  Tue Aug  2 11:12:04 1994
From: john at dcs.rhbnc.ac.uk (john@dcs.rhbnc.ac.uk)
Date: Tue, 02 Aug 94 16:12:04 +0100
Subject: Technical Report Series in Neural and Computational Learning
Message-ID: <8129.9408021512@platon.cs.rhbnc.ac.uk>


The European Community ESPRIT Working Group in Neural and Computational 
       Learning Theory (NeuroCOLT):

---------------------------------------
NeuroCOLT Technical Report NC-TR-94-5:
---------------------------------------
A Weak Version of the Blum, Shub \& Smale Model

by Pascal Koiran

Abstract:
We propose a weak version of the Blum-Shub-Smale model of computation 
over the real numbers. In this weak model only a ``moderate" usage 
of multiplications and divisions is allowed. 
The class of boolean languages recognizable in polynomial
time is shown to be the complexity class P/poly. The main tool is a result
on the existence of small rational points in semi-algebraic sets which is
of independent interest. As an application, we generalize recent results 
of Siegelmann \& Sontag on recurrent neural networks, and of Maass 
on feedforward nets. A preliminary version of this paper was presented at
the 1993 IEEE Symposium on Foundations of Computer Science. Additional
results include:
\begin{itemize}
\item an efficient simulation of order-free real Turing machines by
probabilistic Turing machines in the full Blum-Shub-Smale model;
\item an efficient simulation of arithmetic circuits over the integers by
boolean circuits;
\item the strict inclusion of the real polynomial hierarchy in
weak exponential time.
\end{itemize}


------------------------
The Report NC-TR-94-5 can be accessed and printed as follows 

% ftp cscx.cs.rhbnc.ac.uk  (134.219.200.45)
Name: anonymous
password: your full email address
ftp> cd pub/neurocolt/tech_reports
ftp> binary
ftp> get nc-tr-94-5.ps.Z
ftp> bye
% zcat nc-tr-94-5.ps.Z | lpr -l

Uncompressed versions of the postscript files have also been left for anyone not
having an uncompress facility.

Best wishes
John Shawe-Taylor





From sutton at gte.com  Tue Aug  2 15:54:04 1994
From: sutton at gte.com (Rich Sutton)
Date: Tue, 2 Aug 1994 14:54:04 -0500
Subject: On Step-Size and Bias in TD Learning (Paper Available)
Message-ID: <199408021849.AA05405@ns.gte.com>

The following paper appeared in the proceeding of the 1994 Yale Workshop on
Adaptive and Learning Systems, 1994.  ftp instructions follow.


        ON STEP-SIZE AND BIAS IN TEMPORAL-DIFFERENCE LEARNING

           by Richard S. Sutton and Satinder P. Singh (MIT)
                                       
We present results for three new algorithms for setting the step-size
parameters, alpha and lambda, of temporal-difference learning methods such
as TD(lambda).  The overall task is that of learning to predict the outcome
of an unknown Markov chain based on repeated observations of its state
trajectories. The new algorithms select step-size parameters online in such
a way as to eliminate the bias normally inherent in temporal-difference
methods. We compare our algorithms with conventional Monte Carlo methods.
Monte Carlo methods have a natural way of setting the step size: for each
state s they use a step size of  1/n(s), where n(s) is the number of times
state s has been visited.  We seek and come close to achieving comparable
step-size algorithms for TD(lambda). One new algorithm uses a lambda=1/n(s)
schedule to achieve the same effect as processing a state backwards with
TD(0), but remains completely incremental. Another algorithm uses a lambda
at each time equal to the estimated transition probability of the current
transition. We present empirical results showing improvement in convergence
rate over Monte Carlo methods and conventional TD(lambda).  A limitation of
our results at present is that they apply only to tasks whose state
trajectories do not contain cycles.  

================================================================
unix> ftp envy.cs.umass.edu
 
Name: anonymous
Password: [your ID]
ftp> cd pub/singh
ftp> binary
ftp> get Yale94.ps.Z
ftp> bye
 
unix> uncompress Yale94.ps.Z
unix> [your command to print PostScript file] Yale94.ps
===============================================================
The paper is six pages long.  The file is about 130K.

FTP-host: envy.cs.umass.edu
FTP-filename: /pub/singh/Yale94.ps.Z



From marwan at sedal.su.oz.au  Tue Aug  2 23:12:08 1994
From: marwan at sedal.su.oz.au (Marwan Jabri)
Date: Wed, 3 Aug 1994 13:12:08 +1000
Subject: paper available: Practical Performance and Credit Assignment...
Message-ID: <199408030312.NAA28697@sedal.sedal.su.OZ.AU>

FTP-host: archive.cis.ohio-state.edu
FTP-filename: /pub/neuroprose/jabri.ppap.ps.Z

The file jabri.ppap.ps.Z is now available for
copying from the Neuroprose repository and has been 
submitted for publication:


	Practical Performance and Credit Assignment Efficiency 
	of Analog Multi-layer Perceptron Perturbation Based 
	Training Algorithms

	Marwan A. Jabri

	Systems Engineering and Design Automation Laboratory 
	Sydney University Electrical Engineering
	NSW 2006 Australia
	marwan at sedal.su.oz.au

	SEDAL Technical Report 1-7-94


Abstract
Many algorithms have been recently reported for the training of analog
multi-layer perceptron. Most of these algorithms were evaluated either
from a computational or simulation view point. This paper applies several
of these algorithms to the training of an analog multi-layer perceptron
chip. The advantages and shortcomings of these algorithms in terms of
training and generalisation performance and their capabilities in a
limited precision environment are discussed. Extensive experiments
demonstrate that a trade-off exists between the parallelisation of
perturbations and the efficiency of credit assignment. Two
semi-parallelisation heuristics are presented and are shown to provide
advantages in terms of efficient exploration of the solution space and
fewer credit assignment confusions.


Retrieve this paper by anonymous ftp from:
archive.cis.ohio-state.edu (128.146.8.52)
    in the /pub/neuroprose    directory

The name of the paper in this archive is:
    jabri.ppap.ps.Z          [28 pages]


Sorry, but no hard copies available.

From niranjan at eng.cam.ac.uk  Wed Aug  3 15:07:20 1994
From: niranjan at eng.cam.ac.uk (niranjan@eng.cam.ac.uk)
Date: Wed, 3 Aug 94 15:07:20 BST
Subject: Post Doc job in Cambridge
Message-ID: <18000.9408031407@tulip.eng.cam.ac.uk>

---------------------- JOB  JOB  JOB  JOB -------------------------

    University Of Cambridge, Department of Obstetrics & Gynaecology

	    Post-Doctoral Research Assistant

    Neural Networks in the Quality Assurance in Maternity Care  (QAMC)

Under the terms of a grant recently awarded to the QAMC project by the
Commission of the European Communities (CEC), we expect that we will
soon be able to offer a Post Doctoral Research Position in Cambridge
for the above investigation.


From g.gaskell at psychology.bbk.ac.uk  Thu Aug  4 14:17:00 1994
From: g.gaskell at psychology.bbk.ac.uk (Gareth)
Date: Thu, 4 Aug 94 14:17 BST
Subject: Thesis: Speech Perception & Connectionsim
Message-ID: <mailman.697.1149591313.29955.connectionists@cs.cmu.edu>

(Sorry - ignore the previous message, I sent the wrong
file!)

FTP-host: archive.cis.ohio-state.edu (128.146.8.52)
FTP-filename: /pub/neuroprose/Thesis/gaskell.thesis.ps.Z

A new PhD thesis (150 pages) is now available in the 
neuroprose archive.  The thesis examines the role of 
phonological variation in human speech perception using both 
experimental and connectionist techniques.

Abstract:

The research reported in this thesis examines issues of word 
recognition in human speech perception.  The main aim of the 
research is to assess the effect of regular variation in speech on 
lexical access.  In particular, the effect of a type of neutralising 
phonological variation, assimilation of place of articulation, is 
examined.  This variation occurs regressively across word 
boundaries in connected speech, altering the surface phonetic 
form of the underlying words.  Two methods of investigation 
are used to explore this issue.  Firstly, experiments using cross-
modal priming and phoneme monitoring techniques are used to 
examine the effect of variation on the matching process between 
speech input and lexical form.  Secondly, simulated experiments 
are performed using two computational models of speech 
recognition: TRACE (McClelland & Elman, 1986) and a simple 
recurrent network.

The priming experiments show that the mismatching effects of a 
phonological change on the word-recognition process depend 
on their viability, as defined by phonological constraints.  This 
implies that speech perception involves a process of context-
dependent inference, that recovers the abstract underlying 
representation of speech.  Simulations of these and other 
experiments are then reported using a simple recurrent network 
model of speech perception.  The model accommodates the 
results of the priming studies and predicts that similar 
phonological context effects will occur in non-words.  Two 
phoneme monitoring studies support this prediction, but also 
show interaction between lexical status and viability, implying 
that phonological inference relies on both lexical and 
phonological constraints.  A revision of the network model is 
proposed which learns the mapping from the surface form of 
speech to semantic and phonological representations.

To retrieve the file:

ftp archive.cis.ohio-state.edu
login: anonymous
password:<email address>
ftp> cd /pub/neuroprose/Thesis
ftp> binary
ftp> get gaskell.thesis.ps.Z
ftp> bye
uncompress gaskell.thesis.ps.Z
lpr gaskell.thesis.ps [or whatever you normally do to print]

Gareth Gaskell
Centre for Speech and Language,
Birkbeck College,
London,

g.gaskell at psyc.bbk.ac.uk


From giles at research.nj.nec.com  Thu Aug  4 18:00:26 1994
From: giles at research.nj.nec.com (Lee Giles)
Date: Thu, 4 Aug 94 18:00:26 EDT
Subject: Reprint:LEARNING A CLASS OF LARGE FINITE STATE MACHINES WITH A RECURRENT 
Message-ID: <9408042200.AA18471@fuzzy>

NEURAL NETWORK 

The following reprint is available via the University of Maryland Department 
of Computer Science and the NEC Research Institute archives:

_____________________________________________________________________________



                 Learning a Class of Large Finite State Machines
                      with a Recurrent Neural Network

        UNIVERSITY OF MARYLAND TECHNICAL REPORT UMIACS-TR-94-94 AND CS-TR-3328

                 C. L. Giles[1,2], B. G. Horne[1], T. Lin[1,3] 

      [1] NEC Research Institute, 4 Independence Way, Princeton, NJ  08540
            [2]UMIACS, University of Maryland, College Park, MD 20742
          [3] EE Department, Princeton University, Princeton, NJ 08540

                      {giles,horne,lin}@research.nj.nec.com
                
One of the issues in any learning model is how it scales with problem size.
Neural networks have not been immune to scaling issues. We show that a
dynamically- driven discrete-time recurrent network (DRNN) can learn rather
large grammatical inference problems when the strings of a finite memory
machine (FMM) are encoded as temporal sequences. FMMs are a subclass of
finite state machines which have a finite memory or a finite order of
inputs and outputs. The DRNN that learns the FMM is a neural network that
maps directly from the sequential machine implementation of the FMM. It has
feedback only from the output and not from any hidden units; an example is
the recurrent network of Narendra and Parthasarathy. (FMMs that have zero
order in the feedback of outputs are called definite memory machines and
are analogous to Time-delay or Finite Impulse Response neural networks.)
Due to their topology these DRNNs are as least as powerful as any
sequential machine implementation of a FMM and should be capable of
representing any FMM. We choose to learn ``particular FMMs.\' Specif
ically, these FMMs have a large number of states (simulations are for $256$
and $512$ state FMMs) but have minimal order, relatively small depth and
little logic when the FMM is implemented as a sequential machine.
Simulations for the num ber of training examples versus generalization
performance and FMM extraction size show that the number of training
samples necessary for perfect generalization is less than that necessary to
completely characterize the FMM to be learned. This is in a sense a best
case learning problem since any arbitrarily chosen FMM with a minimal
number of states would have much more order and string depth and most
likely require more logic in its sequential machine implementation.


--------------------------------------------------------------------------

                          FTP INSTRUCTIONS

                unix> ftp cs.umd.edu (128.8.128.8)
                Name: anonymous
                Password: (your_userid at your_site)
                ftp> cd pub/papers/TRs
                ftp> binary
                ftp> get 3328.ps.Z
                ftp> quit
                unix> uncompress 3328.ps.Z

                                 OR

                unix> ftp external.nj.nec.com (138.15.10.100)
                Name: anonymous
                Password: (your_userid at your_site)
                ftp> cd pub/giles/papers
                ftp> binary
                ftp> get large.fsm.ps.Z
                ftp> quit
                unix> uncompress large.fsm.ps.Z

--------------------------------------------------------------------------

--                                 
C. Lee Giles / NEC Research Institute / 4 Independence Way
Princeton, NJ 08540 / 609-951-2642 / Fax 2482
==




From john at dcs.rhbnc.ac.uk  Fri Aug  5 08:00:06 1994
From: john at dcs.rhbnc.ac.uk (john@dcs.rhbnc.ac.uk)
Date: Fri, 05 Aug 94 13:00:06 +0100
Subject: Technical Report Series in Neural and Computational Learning
Message-ID: <10057.9408051200@platon.cs.rhbnc.ac.uk>


The European Community ESPRIT Working Group in Neural and Computational 
       Learning Theory (NeuroCOLT):

Many apologies that the previous technical report NC-TR-94-5 was not available
when first announced. It is now installed. I have also installed an ascii file
entitled `abstracts' which holds an up to date list of the Technical Reports
with their abstracts.

A new technical report is also now available:
---------------------------------------
NeuroCOLT Technical Report NC-TR-94-7:
---------------------------------------
On the power of real Turing machines over binary inputs
by Felipe Cucker, Universitat Pompeu Fabra, Balmes 132, Barcelona 08008, SPAIN
and Dima Grigoriev, Depts. of Comp. Science and Maths, Penn State University,
    University Park, PA 16802, USA

Abstract:

In recent years the study of the complexity of computational problems
involving real numbers has been an increasing research area.  Blum,
Shub and Smale (1989) proposed a computational model ---the real Turing
machine--- for dealing with the above problems was developed.

The aim of this paper is to prove that
$\BP(\PAR)=\;$PSPACE/{\it poly} where $\PAR$ is the class of sets
computed in parallel polynomial time by (ordinary) real Turing machines.
As a consequence we obtain the existence of binary sets that do not
belong to the
Boolean part of $\PAR$ (an extension of the result of Koiran (1994)
since $\PH\subseteq\PAR$) and a separation of complexity classes in
the real setting.

------------------------
The Report NC-TR-94-7 can be accessed and printed as follows (follow the same
procedure for getting the abstracts file)

% ftp cscx.cs.rhbnc.ac.uk  (134.219.200.45)
Name: anonymous
password: your full email address
ftp> cd pub/neurocolt/tech_reports
ftp> binary
ftp> get nc-tr-94-7.ps.Z
ftp> bye
% zcat nc-tr-94-7.ps.Z | lpr -l

Uncompressed versions of the postscript files have also been left for anyone not
having an uncompress facility.

Best wishes
John





From shultz at hebb.psych.mcgill.ca  Fri Aug  5 14:19:36 1994
From: shultz at hebb.psych.mcgill.ca (Tom Shultz)
Date: Fri, 5 Aug 94 14:19:36 EDT
Subject: No subject
Message-ID: <9408051819.AA06064@hebb.psych.mcgill.ca>

Subject: Paper available: Modeling cognitive development on balance 
scale phenomena
Date: 5 Aug. '94

FTP-host: archive.cis.ohio-state.edu
FTP-file: pub/neuroprose/shultz.balance-ml.ps.Z

-------------------------------------------------------------

The following paper has been placed in the Neuroprose archive at 
Ohio State University: 

Modeling cognitive development on balance scale phenomena (33 
pages)

Thomas R. Shultz, Denis Mareschal, & William C. Schmidt
Department of Psychology & McGill Cognitive Science Centre		
McGill University
Montreal, Quebec, Canada H3A 1B1		
shultz at psych.mcgill.ca

Abstract
We used cascade-correlation to model human cognitive development 
on a well studied psychological task, the balance scale. In balance 
scale experiments, the child is asked to predict the outcome of 
placing certain numbers of equal weights at various distances to the 
left or right of a fulcrum. Both stage progressions and information 
salience effects have been found with children on this task. 
Cascade-correlation is a generative connectionist algorithm that 
constructs its own network topology as it learns. Cascade-
correlation networks provided better fits to these human data than 
did previous models, whether rule-based or connectionist. The 
network model was used to generate a variety of novel predictions 
for psychological research.

Keywords: cognitive development, balance scale, connectionist 
learning, cascade-correlation

The paper is published in Machine Learning, 1994, 16, 59-88. 

Instructions for ftp retrieval of this paper are given below.  If you 
are unable to retrieve and print it and therefore wish to receive a 
hardcopy, please send e-mail to shultz at psych.mcgill.ca

Please do not reply directly to this message.

FTP INSTRUCTIONS:

unix> ftp archive.cis.ohio-state.edu (or 128.146.8.52)
    Name: anonymous
    Password: <your e-mail address>
    ftp> cd pub/neuroprose
    ftp> binary
    ftp> get shultz.balance-ml.ps.Z 
    ftp> quit
unix> uncompress shultz.balance-ml.ps.Z 

Thanks to Jordan Pollack for maintaining this archive. 

Tom Shultz

From cohn at psyche.mit.edu  Fri Aug  5 17:00:03 1994
From: cohn at psyche.mit.edu (David Cohn)
Date: Fri, 5 Aug 94 17:00:03 EDT
Subject: Optimal Experiment Design paper available
Message-ID: <9408052100.AA24701@psyche.mit.edu>

----------------------------------------------------------------------

       Neural Network Exploration Using Optimal Experiment Design
                 AI Lab Memo #1491/CBCL Paper #99
 
                          David A. Cohn
               Dept. of Brain and Cognitive Sciences
	       Massachusetts Institute of Technology
                       Cambridge, MA 02139

I consider the question "How should one act when the only goal is to
learn as much as possible?"  Building on the theoretical results of
Fedorov and MacKay, I apply techniques from Optimal Experiment Design
(OED) to guide the query/action selection of a neural network learner.
I demonstrate that these techniques allow the learner to minimize its
generalization error by exploring its domain efficiently and
completely.  I conclude that, while not a panacea, OED-based
query/action has much to offer, especially in domains where its high
computational costs can be tolerated.
----------------------------------------------------------------------

The above paper is a greatly expanded version of one that appeared at
last year's NIPS, and is available by anonymous ftp to:
             publications.ai.mit.edu

in the file: ai-publications/1994/AIM-1491.ps.Z

It is also available from my home page at:
             http://www.ai.mit.edu/people/cohn/cohn.html

I welcome all comments, questions, and (gentle) criticisms.

 -David Cohn				e-mail: cohn at psyche.mit.edu
  Dept. of Brain & Cognitive Science	phone:  (617) 253-8409
  MIT, E10-243
  Cambridge, MA 02139
  http://www.ai.mit.edu/people/cohn/cohn.html

From harnad at Princeton.EDU  Fri Aug  5 20:25:42 1994
From: harnad at Princeton.EDU (Stevan Harnad)
Date: Fri, 5 Aug 94 20:25:42 EDT
Subject: Motor Control (Feldman): BBS Call for Commentators
Message-ID: <9408060025.AA29660@clarity.Princeton.EDU>

    Below is the abstract of a forthcoming target article by:
                A.G. Feldman & M.F. Levin
                        on:
        POSITIONAL FRAMES OF REFERENCE IN MOTOR CONTROL

This article has been accepted for publication in Behavioral and Brain
Sciences (BBS), an international, interdisciplinary journal providing
Open Peer Commentary on important and controversial current research in
the biobehavioral and cognitive sciences.

Commentators must be current BBS Associates or nominated by a current
BBS Associate. To be considered as a commentator for this article, to
suggest other appropriate commentators, or for information about how to
become a BBS Associate, please send email to:

harnad at clarity.princeton.edu  or harnad at pucc.bitnet        or write to:
BBS, 20 Nassau Street, #240, Princeton NJ 08542  [tel: 609-921-7771]

To help us put together a balanced list of commentators, please give
some indication of the aspects of the topic on which you would bring
your areas of expertise to bear if you were selected as a commentator.
An electronic draft of the full text is available for inspection by
anonymous ftp according to the instructions that follow after the abstract.
____________________________________________________________________

    POSITIONAL FRAMES OF REFERENCE IN MOTOR CONTROL: ORIGIN AND USE

        Anatol G. Feldman (1,2,4) & Mindy F. Levin (2,3,4)
    Institute of Biomedical Engineering, University of Montreal (1)
    Research Centre, Rehabilitation Institute of Montreal, H3S 2J4 (2)
    School of Rehabilitation, University of Montreal (3)
    Centre for Research in Neurological Sciences, University of Montreal (4)
                EMAIL:Feldman at ere.umontreal.ca 

    KEYWORDS: motor control, frames of reference, motoneurons, control
    variables, proprioception, kinaesthesis, equilibrium points,
    multi-muscle systems, pointing, synergy, redundancy problem.

    ABSTRACT: A hypothesis about sensorimotor integration (the lambda
    model) is described and applied to movement control and
    kinesthesia.  The nervous system organizes positional frames of
    reference for the sensorimotor apparatus and produces active
    movements by shifting frames in terms of spatial coordinates.
    Kinematic and electromyographic patterns are not programmed but
    emerge from the dynamic interaction of the system's components,
    including external forces, within the designated frame of
    reference. Motoneuronal threshold properties and proprioceptive
    inputs to motoneurons may be important components in the
    physiological mechanism which produces positional frames of
    reference.  The hypothesis that intentional movements are produced
    by shifting the frame of reference is extended to multi-muscle and
    multi-degrees of freedom systems by providing a solution for the
    redundancy problem the allows the control of a joint alone or in
    combination with other joints to produce any desired limb
    configuration and movement trajectory. For each motor behavior, the
    nervous system uses a strategy which minimizes the number of
    changeable control variables and keep sthe parameters of these
    changes invariant. This is illustrated by examples of simulated
    kinematic and electromyographic signals from single- and
    multi-joint arm movements produced by patterns of control
    variables. Empirical support is provided and additional tests are
    suggested. The model is contrasted with others based on the ideas
    of programming of motoneuronal activity, muscle forces, stiffness
    or movement kinematics.

--------------------------------------------------------------
To help you decide whether you would be an appropriate commentator for
this article, an electronic draft is retrievable by anonymous ftp from
princeton.edu according to the instructions below (the filename is
bbs.feldman). Please do not prepare a commentary on this draft.
Just let us know, after having inspected it, what relevant expertise
you feel you would bring to bear on what aspect of the article.
The file is also retrievable using archie, gopher, and World-Wide Web

URLs (Universal Resource Locators):
ftp://princeton.edu/pub/harnad/BBS/
gopher://gopher.princeton.edu/1ftp%3aprinceton.edu%40/pub/harnad/BBS/
http://192.190.21.10/wic/psych.02.html

-------------------------------------------------------------
   To retrieve a file by ftp from an Internet site, type either:
ftp princeton.edu
   or
ftp 128.112.128.1
   When you are asked for your login, type:
anonymous
   Enter password as queried (your password is your actual userid:
   yourlogin at yourhost.whatever.whatever - be sure to include the "@")
cd /pub/harnad/BBS
   To show the available files, type:
ls
   Next, retrieve the file you want with (for example):
get bbs.feldman
   When you have the file(s) you want, type:
quit
----------
Where the above procedure is not available there are two fileservers:
ftpmail at decwrl.dec.com
       and
bitftp at pucc.bitnet
that will do the transfer for you. To one or the
other of them, send the following one line message:
help
for instructions (which will be similar to the above, but will be in
the form of a series of lines in an email message that ftpmail or
bitftp will then execute for you).

JANET users without ftp can instead utilise the file transfer facilities
at sites uk.ac.ft-relay or uk.ac.nsf.sun.  Full details are available on
request.
-------------------------------------------------------------

From bishopc at helios.aston.ac.uk  Mon Aug  8 10:26:53 1994
From: bishopc at helios.aston.ac.uk (bishopc)
Date: Mon, 8 Aug 1994 14:26:53 +0000
Subject: Lectureships in Neural Computing
Message-ID: <29457.9408081326@sun.aston.ac.uk>



-------------------------------------------------------------------


                     Aston University 

               Neural Computing Research Group

     Department of Computer Science and Applied Mathematics



              LECTURESHIPS IN NEURAL COMPUTING
              --------------------------------


Applications are invited for two lectureships commencing in the 
next academic year. Candidates are expected to have excellent
academic qualifications and a proven record of research.
The appointments will be for an initial period of three years,
with the possibility of subsequent renewal or transfer to a
continuing appointment. 

Successful candidates will be expected to make a substantial
contribution to the research activities of the Department in
the area of neural computing. They will also be expected to 
contribute to the undergraduate and postgraduate teaching 
programmes in computer science.

The Neural Computing Research Group currently comprises three
professors, two lecturers, three postdoctoral research fellows and 
ten postgraduate research students. Current research activity
focusses on principled approaches to neural computing, and spans 
a broad spectrum from theoretical foundations to industrial and
commercial applications. These new appointments will further 
strengthen the research activity of this group.

Salaries will be within the lecturer A and B range 
14,756 to 25,735, and exceptionally up to 28,756 (UK pounds).

If you wish to be considered for one of these positions, please 
send a CV and publications list, together with the names of 3 
referees, to:

    Professor Chris Bishop
    Neural Computing Research Group
    Aston University
    Birmingham B4 7ET, U.K.
    Tel: 021 359 3611 ext. 4270
    Fax: 021 333 6215
    e-mail: c.m.bishop at aston.ac.uk



From bishopc at helios.aston.ac.uk  Mon Aug  8 12:36:57 1994
From: bishopc at helios.aston.ac.uk (bishopc)
Date: Mon, 8 Aug 1994 16:36:57 +0000
Subject: Neural Computing Applications Forum
Message-ID: <992.9408081536@sun.aston.ac.uk>


---------------------------------------------------------------

                     NCAF Two-Day Conference:

     PRACTICAL APPLICATIONS AND TECHNIQUES OF NEURAL NETWORKS

                      14 and 15 September 1994

                 Aston University, Birmingham, UK



14 September 1994
-----------------

Tutorial: Introduction to Neural Networks
(How to get started in Neural Computing)
Richard Palmer, ERA Technology

INVITED TALK: From Regularisation to Cheque Verification
Francoise Fogelman, SLIGOS, France

Workshop: Error Bars and Confidence Limits
(Practical Techniques for Assigning Error Bars 
to Network Predictions)

Odyssey: Stocks, Shakespeare and Sunshine:
(or: Useful Tricks with Radial Basis Functions)
David Lowe, Aston University


Social Event: Canal Boat Trip with Wine and Buffet Dinner



15 September 1994
-----------------

Obstacles and Challenges in the Industrial Application of Neural Networks
Inderjit Sandhu, Barclays Bank PLC

Applying Neural Networks to the Analysis of Small Medical Data Sets
Paul Beatty, University Hospital of South Manchester

The Neural Nose
Gareth Jones, Neotronics Ltd

Hand-printed Character Recognition Using Deformable Models
Chris Williams, Aston University

Determination of Ocean Surface Wind Velocities from Satellite Radar Data
Iain Strachan, AEA Technology

Use of Neural Networks to Derive Helicopter Component Load Spectra
Alf Vella, Cranfield University

Working Together: NCAF and the DTI Neural Computing Programme
Bob Wiggins, DTI

The EPSRC Neural Computing Programme
Catherine Barnes and Peter Bates, EPSRC

A Study of Various Neural Network Techniques for Automotive Systems
M Arain, Lucas Advanced Engineering

Predicting Driver Alertness from Steering Behaviour
Kevin Swingler, Stirling University (with Ford UK)


-------------------------------------------------------------------

                 NEURAL COMPUTING APPLICATIONS FORUM


The Neural Computing Applications Forum (NCAF) was formed in 1990
and has since come to provide the principal mechanism for exchange
of ideas and information between academics and industrialists in 
the UK on all aspects of neural networks and their practical 
applications. NCAF organises four 2-day conferences each year, 
which are attended by around 100 participants. It has its own 
international journal `Neural Computing and Applications' which is 
published quarterly by Springer-Verlag, and it produces a quarterly 
newsletter `Networks'. 

Annual membership rates (Pounds Stirling): 

  Company:     250
  Individual:  140
  Associate     90
  Student:      55

Membership includes free registration at all four annual 
conferences, a subscription to the journal `Neural Computing 
and Applications', and a subscription to `Networks'.
Associate membership is intended for those who are unable to 
attend meetings (eg those overseas) but who wish to 
receive the journal and the newsletter, and does not include
registration at the conferences.

For further information: 
   Tel: +44 (0)784 477271
   Fax: +44 (0)784 472879
   email: c.m.bishop at aston.ac.uk


Chris Bishop

(Chairman, NCAF)


-------------------------------------------------------------------- 

  Professor Chris M Bishop           Tel. +44 (0)21 359 3611 x4270
  Neural Computing Research Group    Fax. +44 (0)21 333 6215
  Dept. of Computer Science          c.m.bishop at aston.ac.uk
  Aston University               
  Birmingham B4 7ET, UK

--------------------------------------------------------------------


From harnad at Princeton.EDU  Mon Aug  8 22:32:16 1994
From: harnad at Princeton.EDU (Stevan Harnad)
Date: Mon, 8 Aug 94 22:32:16 EDT
Subject: Subsymbolic Language Processing: PSYC Multiple Book Review
Message-ID: <9408090232.AA12841@clarity.Princeton.EDU>

                CALL FOR BOOK REVIEWERS

Below is the Precis of SUBSYMBOLIC NATURAL LANGUAGE PROCESSING by
Risto Mikkulainen. This book has been selected for multiple review in
PSYCOLOQUY. If you wish to submit a formal book review (see
Instructions following Precis) please write to psyc at pucc.bitnet
indicating what expertise you would bring to bear on reviewing the book
if you were selected to review it. (If you have never reviewed for
PSYCOLOQUY or Behavioral & Brain Sciences before, it would be helpful
if you could also append a copy of your CV to your message.) If you are
selected as one of the reviewers, you will be sent a copy of the book
directly by the publisher (please let us know if you have a copy
already). Reviews may also be submitted without invitation, but all
reviews will be refereed. The author will reply to all accepted
reviews.

----------------------------------------------------------------------
psycoloquy.94.5.46.language-network.1.miikkulainen  Monday  8 Aug 1994
ISSN 1055-0143 (34 paragraphs, 1 fig, 1 note, 16 references, 609 lines)
PSYCOLOQUY is sponsored by the American Psychological Association (APA)
                Copyright 1994 Risto Miikkulainen

                Precis of:
                SUBSYMBOLIC NATURAL LANGUAGE PROCESSING:
                AN INTEGRATED MODEL OF SCRIPTS, LEXICON, AND MEMORY
                Cambridge, MA: MIT Press, 1993
                15 chapters, 403 Pages

                Risto Miikkulainen
                Department of Computer Sciences
                The University of Texas at Austin
                Austin, TX 78712
                risto at cs.utexas.edu

    ABSTRACT: Distributed neural networks have been very successful in
    modeling isolated cognitive phenomena, but complex high-level
    behavior has been amenable only to symbolic artificial intelligence
    techniques. Aiming to bridge this gap, this book describes
    DISCERN, a complete natural language processing system implemented
    entirely at the subsymbolic level. In DISCERN, distributed neural
    network models of parsing, generating, reasoning, lexical
    processing and episodic memory are integrated into a single system
    that learns to read, paraphrase, and answer questions about
    stereotypical narratives. Using DISCERN as an example, a general
    approach to building high-level cognitive models from distributed
    neural networks is introduced, and the special properties of such
    networks are shown to provide insight into human performance. In
    this approach, connectionist networks are not only plausible models
    of isolated cognitive phenomena, but also sufficient constituents
    for generating complex, high-level behavior.

    KEYWORDS: computational modeling, connectionism, distributed neural
    networks, episodic memory, lexicon, natural language processing,
    scripts.

I. MOTIVATION

1. Recently there has been a great deal of excitement in cognitive
science about the subsymbolic (i.e., parallel distributed processing,
or distributed connectionist, or distributed neural network) approach
to natural language processing. Subsymbolic systems seem to capture a
number of intriguing properties of human-like information processing
such as learning from examples, context sensitivity, generalization,
robustness of behavior, and intuitive reasoning. These properties have
been very difficult to model with traditional, symbolic techniques.

2. Within this new paradigm, the central issues are quite different
from (even incompatible with) the traditional issues in symbolic
cognitive science, and the research has proceeded without much in
common with the past. However, the ultimate goal is still the same: to
understand how human cognition is put together. Even if cognitive
science is being built on a new foundation, as can be argued, many of
the results obtained through symbolic research are still valid, and
could be used as a guide for developing subsymbolic models of cognitive
processes.

3. This is where DISCERN, the computer-simulated neural network model
described in this book (Miikkulainen 1993), fits in. DISCERN is a
purely subsymbolic model, but at the high level it consists of modules
and information structures similar to those of symbolic systems, such
as scripts, lexicon, and episodic memory. At the highest level of
cognitive modeling, the symbolic and subsymbolic paradigms have to
address the same basic issues. Outlining a parallel distributed
approach to those issues is the purpose of DISCERN.

4. In more specific terms, DISCERN aims: (1) to demonstrate that
distributed artificial neural networks can be used to build a
large-scale natural language processing system that performs
approximately at the level of symbolic models; (2) to show that several
cognitive phenomena can be explained at the subsymbolic level using the
special properties of these networks; and (3) to identify central
issues in subsymbolic cognitive modeling and to develop well-motivated
techniques to deal with them. To the extent that DISCERN is successful
in these areas, it constitutes a first step towards subsymbolic natural
language processing.

II. THE SCRIPT PROCESSING TASK

5. Scripts (Schank and Abelson, 1977) are schemas of often-encountered,
stereotypic event sequences, such as visiting a restaurant, traveling
by airplane, and shopping at a supermarket. Each script divides further
into tracks, or established minor variations. A script can be
represented as a causal chain of events with a number of open roles.
Script-based understanding means reading a script-based story,
identifying the proper script and track, and filling its roles with the
constituents of the story. Events and role fillers that were not
mentioned in the story but are part of the script can then be inferred.
Understanding is demonstrated by generating an expanded paraphrase of
the original story, and by answering questions about the story.

6. To see what is involved in the task, let us consider an example of
DISCERN input/output behavior. The following input stories are examples
of the fancy-restaurant, plane-travel, and electronics-shopping
tracks:

    John went to MaMaison. John asked the waiter for lobster. John left
    the waiter a big tip.

    John went to LAX. John checked in for a flight to JFK. The plane
    landed at JFK.

    John went to Radio-Shack. John asked the staff questions
    about CD-players. John chose the best CD-player.

7. DISCERN reads the orthographic word symbols sequentially, one at a
time. An internal representation of each story is formed, where all
inferences are made explicit. These representations are stored in the
episodic memory. The system then answers questions about the stories:

    What did John buy at Radio-Shack?
      John bought a CD-player at Radio-Shack.

    Where did John fly to?
      John flew to JFK.

    What did John eat at MaMaison?
      John ate a good lobster.

With the question as a cue, the appropriate story representation is
retrieved from the episodic memory and the answer is generated word by
word. DISCERN also generates full paraphrases of the input stories.
For example, it generates an expanded version of the restaurant story:

    John went to MaMaison. The waiter seated John. John asked the
    waiter for lobster. John ate a good lobster. John paid the waiter.
    John left a big tip. John left MaMaison.

8. The answers and the paraphrase show that DISCERN has made a number
of inferences beyond the original story. For example, it inferred that
John ate the lobster and the lobster tasted good. The inferences are
not based on specific rules but are statistical and learned from
experience. DISCERN has read a number of similar stories in the past
and the unmentioned events and role bindings have occurred in most
cases. They are assumed immediately and automatically upon reading the
story and have become part of the memory of the story. In a similar
fashion, human readers often confuse what was mentioned in the story
with what was only inferred (Bower et al., 1979; Graesser et al.,
1979).

9. A number of issues can be identified from the above examples.
Specifically, DISCERN has to (1) make statistical, script-based
inferences and account for learning them from experience; (2) store
items in the episodic memory in a single presentation and retrieve them
with a partial cue; (3) develop a meaningful organization for the
episodic memory, based on the stories it reads; (4) represent meanings
of words, sentences, and stories internally; and (5) organize a lexicon
of symbol and concept representations based on examples of how words
are used in the language and form a many-to-many mapping between them.
Script processing constitutes a good framework for studying these
issues, and a good domain for developing an approach towards the goals
outlined above.

III. APPROACH

10. Parallel distributed processing models typically have very little
internal structure. They produce the statistically most likely answer
given the input conditions in a process that is opaque to the external
observer. This is well suited to the modeling of isolated low-level
tasks, such as learning past tense forms of verbs (Rumelhart and
McClelland, 1986) or word pronunciation (Sejnowski and Rosenberg,
1987). Given the success of such models, a possible approach to
higher-level cognitive modeling would be to construct the system from
several submodules that work together to produce the higher-level
behavior.

11. In DISCERN, the immediate goal is to build a complete, integrated
system that performs well in the script processing task. In this sense,
DISCERN is very similar to traditional models in artificial
intelligence. However, DISCERN also aims to show how certain parts of
human cognition could actually be built. The components of DISCERN were
designed as independent cognitive models that can account for
interesting language processing and memory phenomena, many of which are
not even required in the DISCERN task. Combining these models into a
single, working system is one way of validating them. In DISCERN, the
components are not just models of isolated cognitive phenomena; they
are sufficient constituents for generating complex high-level
behavior.

IV. THE DISCERN MODEL

12. DISCERN can be divided into parsing, generating, question
answering, and memory subsystems, each with two modules (figure 1).
Each module is trained in its task separately and in parallel. During
performance, the modules form a network of networks, each feeding its
output to the input of another module.

                       Input text     Output text
                              |         ^
                              V         |
  =================        =================        =================
   Sentence Parser <-------     Lexicon     <------- Sentence Gener.
  =================        =================        =================
     |         |                                           ^     ^
     |         |                                           |     |
     |         +-------+-----------------------+           |     |
     |                 |                       |           |     |
     |                 V                       V           |     |
     |         =================       =================   |     |
     |             Cue Former           Answer Producer ---+     |
     |         =================       =================         |
     |                       |           ^                       |
     |                       |           |                       |
     V                       V           |                       |
  =================        =================        =================
    Story Parser   -------> Episodic Memory -------> Story Generator
  =================        =================        =================

                      Figure 1: The DISCERN Model.

13. The sentence parser reads the input words one at a time and forms
a representation of each sentence. The story parser combines the
sequence of sentences into an internal representation of the story,
which is then stored in the episodic memory. The story generator
receives the internal representation and generates the sentences of the
paraphrase one at a time. The sentence generator outputs the sequence
of words for each sentence. The cue former receives a question
representation, built by the sentence parser, and forms a cue pattern
for the episodic memory, which returns the appropriate story
representation. The answer producer receives the question and the story
and generates an answer representation, which is output word by word by
the sentence generator. The architecture and behavior of each of these
modules in isolation is outlined below.

V. LEXICON

14. The input and output of DISCERN consist of distributed
representations for orthographic word symbols (also called lexical
words). Internally, DISCERN processes semantic concept representations
(semantic words). Both the lexical and semantic words are represented
distributively as vectors of gray-scale values between 0.0 and 1.0. The
lexical representations are based on the visual patterns of characters
that make up the written word; they remain fixed throughout the
training and performance of DISCERN. The semantic representations stand
for distinct meanings and are developed automatically by the system
while it is learning the processing task.

15. The lexicon stores the lexical and semantic representations and
translates between them. It is implemented as two feature maps
(Kohonen, 1989), one lexical and the other semantic. Words whose
lexical forms are similar, such as "LINE" and "LIKE", are represented
by nearby units in the lexical map. In the semantic map, words with
similar semantic content, such as "John" and "Mary", or "Leone's" and
"MaMaison" are mapped near each other. There is a dense set of
associative interconnections between the two maps. A localized activity
pattern representing a word in one map will cause a localized activity
pattern to form in the other map, representing the same word. The
output representation is then obtained from the weight vector of the
most highly active unit. The lexicon thus transforms a lexical input
vector into a semantic output vector and vice versa. Both maps and the
associative connections between them are organized simultaneously,
based on examples of co-occurring symbols and meanings.

16. The lexicon architecture facilitates interesting behavior.
Localized damage to the semantic map results in category-specific
lexical deficits similar to human aphasia (Caramazza, 1988; McCarthy and
Warrington, 1990). For example, the system selectively loses access to
restaurant names, or animate words, when that part of the map is
damaged. Dyslexic performance errors can also be modeled. If the
performance is degraded, for example, by adding noise to the
connections, parsing and generation errors that occur are quite similar
to those observed in human deep dyslexia (Coltheart et al., 1988). For
example, the system may confuse "Leone's" with "MaMaison", or "LINE"
with "LIKE", because they are nearby in the map and share similar
associative connections.

VI. FGREP PROCESSING MODULES

17. Processing in DISCERN is carried out by hierarchically organized
pattern-transformation networks. Each module performs a specific
subtask, such as parsing a sentence or generating an answer to a
question. All these networks have the same basic architecture: they are
three-layer, simple-recurrent backpropagation networks (Elman, 1990),
with the extension called FGREP that allows them to develop distributed
representations for their input/output words.

18. The network learns the processing task by adapting the connection
weights according to the standard on-line backpropagation procedure
(Rumelhart et al., 1986, pp. 327-329). The error signal is propagated
to the input layer, and the current input representations are modified
as if they were an extra layer of weights. The modified representation
vectors are put back in the lexicon, replacing the old representations.
Next time the same words occur in the input or output, their new
representations are used to form the input/output patterns for the
network. In FGREP, therefore, the required mappings change as the
representations evolve, and backpropagation is shooting at a moving
target.

19. The representations that result from this process have a number of
useful properties for cognitive modeling. (1) Since they adapt to the
error signal, they end up coding information most crucial to the task.
Representations for words that are used in similar ways in the examples
become similar. Thus, these profiles of continuous activity values can
be claimed to code the meanings of the words as well. (2) As a result,
the system never has to process very novel input patterns, because
generalization has already been done in the representations. (3) The
representation of a word is determined by all the contexts in which
that word has been encountered; consequently, it is also a
representation of all those contexts. Expectations emerge automatically
and cumulatively from the input word representations. (4) Single
representation components do not usually stand for identifiable
semantic features. Instead, the representation is holographic: word
categories can often be recovered from the values of single
components. (5) Holography makes the system very robust against noise
and damage. Performance degrades approximately linearly as
representation components become defective or inaccurate.

VII. EPISODIC MEMORY

20. The episodic memory in DISCERN consists of a hierarchical pyramid
of feature maps organized according to the taxonomy of script-based
stories. The highest level of the hierarchy is a single feature map
that lays out the different script classes. Beneath each unit of this
map there is another feature map that lays out the tracks within the
particular script. The different role bindings within each track are
separated at the bottom level. The map hierarchy receives a story
representation vector as its input and classifies it as an instance of
a particular script, track, and role binding. The hierarchy thereby
provides a unique memory representation for each script-based story as
the maximally responding units in the feature maps at the three
levels.

21. Whereas the top and the middle level in the hierarchy only serve as
classifiers, selecting the appropriate track and role-binding map for
each input, at the bottom level a permanent trace of the story must
also be created. The role-binding maps are trace feature maps, with
modifiable lateral connections. When the story representation vector is
presented to a role-binding map, a localized activity pattern forms as
a response. Each lateral connection to a unit with higher activity is
made excitatory, while a connection to a unit with lower activity is
made inhibitory. The units within the response now "point" towards the
unit with highest activity, permanently encoding that the story was
mapped at that location.

22. A story is retrieved from the episodic memory by giving it a
partial story representation as a cue. Unless the cue is highly
deficient, the map hierarchy is able to recognize it as an instance of
the correct script and track and form a partial cue for the
role-binding map. The trace feature map mechanism then completes the
role binding. The initial response of the map is again a localized
activity pattern; because the map is topological, it is likely to be
located somewhere near the stored trace. If the cue is close enough,
the lateral connections pull the activity to the center of the stored
trace. The complete story representation is retrieved from the weight
vectors of the maximally responding units at the script, track, and
role-binding levels.

23. Hierarchical feature maps have a number of properties that make
them useful for memory organization: (1) The organization is formed in
an unsupervised manner, extracting it from the input experience of the
system. (2) The resulting order reflects the properties of the data,
the hierarchy corresponding to the levels of variation, and the maps
laying out the similarities at each level. (3) By dividing the data
first into major categories and gradually making finer distinctions
lower in the hierarchy, the most salient components of the input data
are singled out and more resources are allocated for representing them
accurately. (4) Because the representation is based on salient
differences in the data, the classification is very robust, and usually
correct even if the input is noisy or incomplete. (5) Because the
memory is based on classifying the similarities and storing the
differences, retrieval becomes a reconstructive process (Kolodner, 1984;
Williams and Hollan, 1981) similar to human memory.

24. The trace feature map exhibits interesting memory effects that
result from interactions between traces. Later traces capture units
from earlier ones, making later traces more likely to be retrieved.
The extent of the traces determines memory capacity. The smaller the
traces, the more of them will fit in the map, but more accurate cues
are required to retrieve them. If the memory capacity is exceeded,
older traces will be selectively replaced by newer ones. Traces that
are unique, that is, located in a sparse area of the map, are not
affected, no matter how old they are. Similar effects are common in
human long-term memory (Baddeley, 1976; Postman, 1971).

VIII. DISCERN HIGH-LEVEL BEHAVIOR

25. DISCERN is more than just a collection of individual cognitive
models. Interesting behavior results from the interaction of the
components in a complete story-processing system.

26. DISCERN was trained and tested with an artificially generated
corpus of script-based stories consisting of three scripts with three
tracks and three open roles each. The complete DISCERN system performs
very well: at the output, about 98 percent of the words are correct.
This is rather remarkable for a chain of networks that is 9 modules
long and consists of several different types of modules.

27. A modular neural network system can only operate if it is stable,
that is, if small deviations from the normal flow of information are
automatically corrected. It turns out that DISCERN has several built-in
safeguards against minor inaccuracies and noise. The semantic
representations are distributed and redundant, and inaccuracies in the
output of one module are cleaned up by the module that uses the
output. The memory modules clean up by categorical processing: a noisy
input is recognized as a representative of an established class and
replaced by the correct representation of that class. As a result,
small deviations do not throw the system off course, but rather the
system filters out the errors and returns to the normal course of
processing, which is an essential requirement for building robust
cognitive models.

28. DISCERN also demonstrates strong script-based inferencing. Even
when the input story is incomplete, consisting of only a few main
events, DISCERN can usually form an accurate internal representation
of it. DISCERN was trained to form complete story representations
from the first sentence on, and because the stories are stereotypical,
missing sentences have little effect on the parsing process. Once the
story representation has been formed, DISCERN performs as if the script
had been fully instantiated. Questions about missing events and
role-bindings are answered as if they were part of the original story.
If events occurred in an unusual order, they are recalled in the
stereotypical order in the paraphrase. If there is not enough
information to fill a role, the most likely filler is selected and
maintained throughout the paraphrase generation. Such behavior
automatically results from the modular architecture of DISCERN and is
consistent with experimental observations on how people remember
stories of familiar event sequences (Bower et al., 1979; Graesser et
al., 1979).

29. In general, given the information in the question, DISCERN recalls
the story that best matches it in the memory. An interesting issue
is: what happens when DISCERN is asked a question that is inaccurate
or ambiguous, that is, one that does not uniquely specify a story? For
example, DISCERN might have read a story about John eating lobster at
MaMaison, and then about Mary doing the same at Leone's, and the
question could be "Who ate lobster?" Because later traces are more
prominent in the memory, DISCERN is more likely to retrieve the
Mary-at-Leone's story in this case. The earlier story is still in the
memory, but to recall it, more details need to be specified in the
question, such as `Who ate lobster at MaMaison?" Similarly, DISCERN can
robustly retrieve a story even if the question is slightly inaccurate.
When asked "How did John like the steak at MaMaison?", DISCERN
generates the answer "John thought lobster was good at MaMaison",
ignoring the inaccuracy in the question, because the cue is still close
enough to the stored trace. DISCERN does recognize, though, when a
question is too different from anything in the memory, and should not
be answered. For "Who ate at McDonald's?", the cue vector is not close
to any trace, the memory does not settle, and nothing is retrieved.
Note that these mechanisms were not explicitly built into DISCERN,
but they emerge automatically from the physical layout of the
architecture and representations.

IX. DISCUSSION

30. There is an important distinction between scripts (or more
generally, schemas) in symbolic systems, and scripts in subsymbolic
models such as DISCERN. In the symbolic approach, a script is stored
in memory as a separate, exact knowledge structure, coded by the
knowledge engineer. The script has to be instantiated by searching the
schema memory sequentially for a structure that matches the input.
After instantiation, the script is active in the memory and later
inputs are interpreted primarily in terms of this script. Deviations
are easy to recognize and can be taken care of with special
mechanisms.

31. In the subsymbolic approach, schemas are based on statistical
properties of the training examples, extracted automatically during
training. The resulting knowledge structures do not have explicit
representations. For example, a script exists in a neural network only
as statistical correlations coded in the weights. Every input is
automatically matched to every correlation in parallel. There is no
all-or-none instantiation of a particular knowledge structure. The
strongest, most probable correlations will dominate, depending on how
well they match the input, but all of them are simultaneously active at
all times. Regularities that make up scripts can be particularly well
captured by such correlations, making script-based inference a good
domain for the subsymbolic approach. Generalization and graceful
degradation give rise to inferencing that is intuitive, immediate, and
occurs without conscious control, as is script-based inference in
humans. On the other hand, it is very difficult to recognize deviations
from the script and to initiate exception-processing when the automatic
mechanisms fail. Such sequential reasoning would require intervention
of a high-level "conscious" monitor, which has yet to be built in the
connectionist framework.

X. CONCLUSION

32. The main conclusion from DISCERN is that building subsymbolic
models is a feasible approach to understanding mechanisms underlying
natural language processing. DISCERN shows how several cognitive
phenomena may result from subsymbolic mechanisms. Learning word
meanings, script processing, and episodic memory organization are based
on self-organization and gradient-descent in error in this model.
Script-based inferences, expectations, and defaults automatically
result from generalization and graceful degradation. Several types of
performance errors in role binding, episodic memory, and lexical access
emerge from the physical organization of the system. Perhaps most
significantly, DISCERN shows how individual connectionist models can be
combined into a large, integrated system that demonstrates that these
models are sufficient constituents for generating sequential, symbolic,
high-level behavior.

33. Although processing simple script instantiations is a start, there
is a long way to go before subsymbolic models will rival the best
symbolic cognitive models. For example, in story understanding, symbolic
systems have been developed that analyze realistic stories in depth,
based on higher-level knowledge structures such as goals, plans,
themes, affects, beliefs, argument structures, plots, and morals. In
designing subsymbolic models that would do that, we are faced with two
major challenges: (1) how to implement connectionist control of
high-level processing strategies (making it possible to model processes
more sophisticated than a series of reflex responses), and (2) how to
represent and learn abstractions (making it possible to process
information at a higher level than correlations in the raw input
data). Progress in these areas would constitute a major step towards
extending the capabilities of subsymbolic natural language processing
models beyond those of DISCERN.

XI. NOTE

34. Software for the DISCERN system is available through anonymous ftp
from cs.utexas.edu:pub/neural-nets/discern. An X11 graphics demo,
showing DISCERN in processing the example stories discussed in the
book, can be run remotely under the World Wide Web at
http://www.cs.utexas.edu/~risto/discern.html, or by telnet with "telnet
cascais.utexas.edu 30000".

XII. TABLE OF CONTENTS

PART I Overview
1 Introduction
2 Background
3 Overview of DISCERN

PART II Processing Mechanisms
4 Backpropagation Networks
5 Developing Representations in FGREP Modules
6 Building from FGREP Modules

PART III Memory Mechanisms
7 Self-Organizing Feature Maps
8 Episodic Memory Organization: Hierarchical Feature Maps
9 Episodic Memory Storage and Retrieval: Trace Feature Maps
10 Lexicon

PART IV Evaluation
11 Behavior of the Complete Model
12 Discussion
13 Comparison to Related Work
14 Extensions and Future Work
15 Conclusions

APPENDICES
A Story Data
B Implementation Details
C Instructions for Obtaining the DISCERN Software

XIII. REFERENCES

Baddeley, A.D. (1976) The Psychology of Memory. New York: Basic Books.

Bower, G.H., Black, J.B. and Turner, T.J. (1979) Scripts in memory for
text. Cognitive Psychology, 11:177-220.

Caramazza, A. (1988) Some aspects of language processing revealed
through the analysis of acquired aphasia: The lexical system. Annual
Review of Neuroscience, 11:395-421.

Coltheart, M., Patterson, K. and Marshall, J.C., editors (1988) Deep
Dyslexia. London; Boston: Routledge and Kegan Paul. Second edition.

Elman, J.L. (1990) Finding structure in time. Cognitive Science,
14:179-211.

Graesser, A.C., Gordon, S.E. and Sawyer, J.D. (1979) Recognition memory
for typical and atypical actions in scripted activities: Tests for the
script pointer+tag hypothesis. Journal of Verbal Learning and Verbal
Behavior, 18:319-332.

Kohonen, T. (1989) Self-Organization and Associative Memory. Berlin;
Heidelberg; New York: Springer. Third edition.

Kolodner, J.L. (1984) Retrieval and Organizational Strategies in
Conceptual Memory: A Computer Model. Hillsdale, NJ: Erlbaum.

Miikkulainen, R. (1993) Subsymbolic Natural Language Processing:  an
Integrated Model of Scripts, Lexicon, and Memory. Cambridge MA: MIT.

McCarthy, R.A. and Warrington, E.K. (1990) Cognitive Neuropsychology: A
Clinical Introduction. New York: Academic Press.

Postman, L. (1971) Transfer, interference and forgetting. In Kling,
J.W., and Riggs, L.A., editors, Woodworth and Schlosberg's Experimental
Psychology, 1019-1132. New York: Holt, Rinehart and Winston. Third
edition.

Rumelhart, D.E. and McClelland, J.L. (1986) On learning past tenses of
English verbs. In Rumelhart, D.E., and McClelland, J.L., editors,
Parallel Distributed Processing: Explorations in the Microstructure of
Cognition, Volume 2, 216--271. Cambridge, MA: MIT Press.

Rumelhart, D.E., Hinton, G.E. and Williams, R.J. (1986) Learning
internal representations by error propagation. In Rumelhart, D.E. and
McClelland, J.L., editors, Parallel Distributed Processing:
Explorations in the Microstructure of Cognition, Volume 1, 318-362.
Cambridge, MA: MIT Press.

Sejnowski, T. J., and Rosenberg, C. R. (1987) Parallel networks that
learn to pronounce English text. Complex Systems, 1:145--168.

Schank, R.C. and Abelson, R.P. (1977) Scripts, Plans, Goals, and
Understanding: An Inquiry into Human Knowledge Structures. Hillsdale,
NJ: Erlbaum.

Williams, M.D. and Hollan, J.D. (1981) The process of retrieval from
very long-term memory. Cognitive Science, 5:87--119.

--------------------------------------------------------------------

        PSYCOLOQUY Book Review Instructions

The PSYCOLOQUY book review procedure is very similar to the commentary
procedure except that it is the book itself, not a target article, that
is under review. (The Precis summarizing the book is intended to permit
PSYCOLOQUY readers who have not read the book to assess the exchange,
but the reviews should address the book, not primarily the Precis.)

Note that as multiple reviews will be co-appearing, you need only
comment on the aspects of the book relevant to your own specialty and
interests, not necessarily the book in its entirety. Any substantive
comments and criticism -- including points calling for a detailed and
substantive response from the author -- are appropriate. Hence,
investigators who have already reviewed or intend to review this book
elsewhere are still encouraged to submit a PSYCOLOQUY review specifically
written with this specialized multilateral review-and-response feature
in mind.

1.  Before preparing your review, please read carefully
    the Instructions for Authors and Commentators and examine
    recent numbers of PSYCOLOQUY.

2.  Reviews should not exceed 500 lines. Where judged necessary
    by the Editor, reviews will be formally refereed.

3.  Please provide a title for your review. As many
    commentators will address the same general topic, your
    title should be a distinctive one that reflects the gist
    of your specific contribution and is suitable for the
    kind of keyword indexing used in modern bibliographic
    retrieval systems. Each review should also have a brief
    (~50-60 word) Abstract

4.  All paragraphs should be numbered consecutively. Line length
    should not exceed 72 characters.  The review should begin with
    the title, your name and full institutional address (including zip
    code) and email address.  References must be prepared in accordance
    with the examples given in the Instructions.  Please read the
    sections of the Instruction for Authors concerning style,

    INSTRUCTIONS FOR PSYCOLOQUY AUTHORS AND COMMENTATORS

PSYCOLOQUY is a refereed electronic journal (ISSN 1055-0143) sponsored
on an experimental basis by the American Psychological Association
and currently estimated to reach a readership of 40,000. PSYCOLOQUY
publishes brief reports of new ideas and findings on which the author
wishes to solicit rapid peer feedback, international and
interdisciplinary ("Scholarly Skywriting"), in all areas of psychology
and its related fields (biobehavioral science, cognitive science,
neuroscience, social science, etc.). All contributions are refereed.

Target article length should normally not exceed 500 lines [c. 4500 words].
Commentaries and responses should not exceed 200 lines [c. 1800 words].

All target articles, commentaries and responses must have (1) a short
abstract (up to 100 words for target articles, shorter for commentaries
and responses), (2) an indexable title, (3) the authors' full name(s)
and institutional address(es).

In addition, for target articles only: (4) 6-8 indexable keywords,
(5) a separate statement of the authors' rationale for soliciting
commentary (e.g., why would commentary be useful and of interest to the
field? what kind of commentary do you expect to elicit?) and
(6) a list of potential commentators (with their email addresses).

All paragraphs should be numbered in articles, commentaries and
responses (see format of already published articles in the PSYCOLOQUY
archive; line length should be < 80 characters, no hyphenation).

It is strongly recommended that all figures be designed so as to be
screen-readable ascii. If this is not possible, the provisional
solution is the less desirable hybrid one of submitting them as
postscript files (or in some other universally available format) to be
printed out locally by readers to supplement the screen-readable text
of the article.

PSYCOLOQUY also publishes multiple reviews of books in any of the above
fields; these should normally be the same length as commentaries, but
longer reviews will be considered as well. Book authors should submit a
500-line self-contained Precis of their book, in the format of a target
article; if accepted, this will be published in PSYCOLOQUY together
with a formal Call for Reviews (of the book, not the Precis). The
author's publisher must agree in advance to furnish review copies to the
reviewers selected.

Authors of accepted manuscripts assign to PSYCOLOQUY the right to
publish and distribute their text electronically and to archive and
make it permanently retrievable electronically, but they retain the
copyright, and after it has appeared in PSYCOLOQUY authors may
republish their text in any way they wish -- electronic or print -- as
long as they clearly acknowledge PSYCOLOQUY as its original locus of
publication. However, except in very special cases, agreed upon in
advance, contributions that have already been published or are being
considered for publication elsewhere are not eligible to be considered
for publication in PSYCOLOQUY,

Please submit all material to psyc at pucc.bitnet or psyc at pucc.princeton.edu
Anonymous ftp archive is DIRECTORY pub/harnad/Psycoloquy HOST princeton.edu


From fellous at selforg.usc.edu  Tue Aug  9 19:05:48 1994
From: fellous at selforg.usc.edu (Jean-Marc Fellous)
Date: Tue, 9 Aug 1994 16:05:48 -0700
Subject: Position in Bochum
Message-ID: <199408092305.QAA08804@selforg.usc.edu>



Please post the following Job announcement:


-----------------------

The following announcement is related to a C3 assistant professor at the
University of Bochum (Germany) in the field of neuro-computation.
It is intentionally kept in its original German form, since knowledge of German
is prefered but not required (However, non-German speaking applicants are expected
to eventually learn German ...).



              Ruhr-Universit"at Bochum

Am Institut f"ur Neuroinformatik ist eine C3-Professur
"Neuroinformatik" zu besetzen.

Das Institut ist eine zentrale wissenschaftliche Einrichtung der
Universit"at mit den Abteilungen Systembiophysik und Theoretische
Biologie.  Arbeitsschwerpunkte sind Prinzipien der Selbst-organisation
und Informationsverarbeitung in neuronaler Architektur.  Die Position
schlie{\ss}t die Mitwirkung der kollegialen Leitung des Instituts ein.
Zu den Lehraufgaben geh"oren Vorlesungen "uber Neuronale Netze sowie
"uber technisch nutzbare Organisationsprinzipien biologischer
neuronaler Systeme.

Es wird erwartet, da{\ss} die zu berufende Pers"onlichkeit  
mindestens in einem der folgenden Gebiete wissenschaftlich  
ausgewiesen ist:
	- Analyse und Anwendung biologischer neuronaler 
          Organisationsprinzipien
	- K"unstliche Neuronale Netze
	- Entwurf von Systemen in neuronaler Architektur
	- Probleme der Selbstorganisation

Neben Erfahrungen im theoretischen Bereich (Systemtheorie,  
nicht-lineare Dynamik) wird Interesse an  
anwendungsorientierten Problemen vorausgesetzt. Eine  
Kooperation mit dem Zentrum f"ur Neuroinformatik in Bochum,  
das bevorzugt anwendungsorientierte Probleme bearbeitet,  
ist m"oglich.

Die Ruhr-Universit"at Bochum bem"uht sich um die F"orderung von Frauen
in Forschung und Lehre.  Schwerbehinderte werden bei gleicher
Qualifikation bevorzugt.

Ihre schriftliche Bewerbung mit den "ublichen Unterlagen  
richten Sie bitte an den Rektor der Ruhr-Universit"at  
Bochum, 44780 Bochum, Germany.



----- End Included Message -----


From druck at afit.af.mil  Wed Aug 10 15:40:12 1994
From: druck at afit.af.mil (Dennis W. Ruck)
Date: Wed, 10 Aug 94 15:40:12 -0400
Subject: CFP: SPIE Applications and Science of Artificial Neural Networks VI
Message-ID: <9408101940.AA06900@gandalf.afit.af.mil>

                        CALL FOR PAPERS

                       SPIE Conference on
  Applications and Science of Artificial Neural Networks VI
 	     

	                Orlando, Florida
		        17-21 April 1995    


You are invited to submit a paper to the SPIE conference Applications and  
Science of Artificial Neural Networks VI. This conference will be held in  
conjunction with nearly 40 other conferences on topics in object recognition,  
aerospace sensing, and photonics. See below for a complete list. 


--------------------------------------------------------------
              Announcement and Call for Papers
	                   for
  Applications and Science of Artificial Neural Networks VI
--------------------------------------------------------------
Conference Chairs: Steven K. Rogers, Dennis W. Ruck, Air Force
Institute of Technology

Program Committee: Stanley C. Ahalt, The Ohio State Univ.; James C.
Bezdek, Univ. of West Florida; Joe R. Brown, Microelectronics and
Computer Technology Corp.; Lee A. Feldkamp, Ford Motor Co.; Michael
Georgiopoulos, Univ. of Central Florida; Joydeep Ghosh, Univ. of
Texas/Austin; Charles W. Glover, Oak Ridge National Lab.; John B.
Hampshire, II, Jet Propulsion Lab.; Richard P. Lippmann, MIT
Lincoln Lab.; Harley R. Myler, Univ. of Central Florida; Mary Lou
Padgett, Auburn Univ.; Kevin L. Priddy, Accurate Automation Corp.;
Gintaras V. Puskorius, Ford Motor Co.;
Donald F. Specht, Lockheed Palo Alto Research Lab.; Gregory L.
Tarr, Air Force Phillips Lab.; Gary Whittington, Univ. of Aberdeen
(UK); Rodney G. Winter, Dept. of Defense

The focus of this conference is on real-world applications of artificial neural  
networks and on recent theoretical developments applicable to current  
applications. The goal of this conference is to provide a forum for interaction  
between researchers and industrial/government agencies with information  
processing requirements. Papers that investigate advantages/disadvantages of  
artificial neural networks in specific real-world applications will be  
presented. Papers that clearly state existing problems in information  
processing that could potentially be solved by artificial neural networks will  
also be considered.

Sessions will concentrate on:

* innovative applications of artificial neural networks to solve
  real-world problems
* comparative performance in applications of target recognition,
  object recognition, speech processing, speaker identification,
  cochannel processing, signal processing in realistic
  environments, robotics, process control, and image processing
* demonstrations of properties and limitations of existing or new
  artificial neural networks as shown by or related to an
  application
* environments for artificial neural networks development and
  implementation with specific applications used to demonstrate
  features of the systems
* hardware implementation technologies that are general purpose or
  application specific
* knowledge acquisition and representation
* biologically inspired visual representation techniques
* decision support systems
* artificial life
* cognitive science
* hybrid systems (fuzzy, neural, genetic)
* neurobiology
* optimization
* sensation and perception
* system identification
* financial applications
* time series analysis and prediction
* pattern recognition
* medical applications
* intelligent control
* robotics.

------------------------------------------------------------------
                    List of Conferences
------------------------------------------------------------------

The following conferences will all be held in Orlando, Florida 17-21 April 1995  
at the Marriott Orlando World Center:

  1.  Public Safety/Law Enforcement Technology
  2.  Photonics for Space Environments III
  3.  Space Environmental, Legal, and Safety Issues
  4.  Imaging Spectrometry
  5.  Commercialization of High-Resolution Satellite Imagery
      for Dual-Use Applications
  6.  Infrared Detectors and Instrumentation for Astronomy
  7.  Spaceborne Interferometry II
  8.  Space Telescopes and Instruments III
  9.  Fiber Optics in Astronomical Applictions
 10.  Telescope Control Systems
 11.  Distributed Interactive Simulation (Critical Reviews)
 12.  Helmet- and Head-Mounted Displays and Symbology Design
      Requirements II
 13.  Cockpit Displays II
 14.  Space Guidance, Control, and Tracking II
 15.  Synthetic Vision for Vehicle Guidance and Control
 16.  Acquisition, Tracking, and Pointing IX
 17.  Applied Laser Radar Technology II
 18.  Air Traffic Control Technologies
 19.  Technologies for Advanced Land Combat (Critical Reviews)
      Part I: Rapid Force Projection Initiative
      Part II: Advanced Vehicle Technologies
      Part III: Information Sciences for Digitizing the Battlefield
 20.  Detection Technologies for Mines and Minelike Targets
 21.  Targets and Backgrounds: Characterization and Representation
 22.  Atmospheric Propagation and Remote Sensing IV
 23.  Tactical Control Technologies
 24.  Test and Evaluation of Defense-Related Infrared Detectors
      and Arrays
 25.  Infrared Imaging Systems: Design, Analysis, Modeling,
      and Testing VI
 26.  Signal Processing, Sensor Fusion, and Target Recognition IV
 27.  Algorithms for Synthetic Aperture Radar Imagery II
 28.  Integration Photogrammetric Techniques with Scene Analysis
      and Machine Vision II
 29.  Automatic Object Recognition V
 30.  Smart Infrared Focal Plane Arrays and Technology
 31.  Transition of Optical Processors into Systems 1995
 32.  Optical Pattern Recognition VI
 33.  Visual Information Processing IV
 34.  Applications and Science of Artificial Neural Networks VI
 35.  Applications of Fuzzy Logic Technology II
 36.  Thermosense XVII: An International Conference on Thermal
      Sensing and Imaging Diagnostic Applications
 37.  Flat Panel Displays for Defense Applications
      (Critical Reviews)
 38.  Digital Signal Processing Technology (Critical Reviews)


------------------------------------------------------------------
                    General Information
------------------------------------------------------------------
SPIE's International Symposium on
Aerospace/Defense Sensing and Dual-Use Photonics
17-21 April 1995
Marriott's Orlando World Center
Resort and Convention Center
Orlando, Florida USA

SPIE's 1995 Aerospace/Defense Sensing and Dual-Use Photonics
Symposium will be held at:
Marriott's Orlando World Center Hotel
8701 World Center Drive
Orlando, Florida 32821-6398
Phone: 407/239-4200 or 800/621-0638 (outside Florida)
Fax: 407/239-5958

Accommodations
--------------
SPIE will reserve a block of rooms for attendees at the Marriott
Orlando World Center Hotel. Room rates at the Marriott will be $129
single and $142 double plus tax. Alternate hotels in the immediate
area will also be available. Information concerning hotels and
prices will be announced in the advance program.


Advance Technical Program
-------------------------
The comprehensive Advance Technical Program for this symposium will
list conferences, paper titles and authors in order of
presentation, educational short courses schedule including course
descriptions and instructor biographies, and an outline of all
planned special events. Call SPIE at 206/676-3290 (Pacific Time) to
request that a copy be sent to you when it becomes available in
January 1995.


Conference Registration
-----------------------
The following registration fees for SPIE's International Symposium
on Aerospace/Defense Sensing and Dual-Use Photonics are preliminary
and included to assist you in planning.

Conference Fees without Proceedings       Member      Nonmember
  Attendee Full Conference.................$360.........$420
  One day...................................160..........190
  Author Full Conference*...................325..........385
  Author One Day*...........................160..........190
  Student....................................85...........95

* Author fee includes a proceedings

Short Course Fees

                              Member    Nonmember
Half-day course (3.5 hr).......$145.......$170
Full-day course (6.5 hr)........265........310
Two-day course (12 hr)..........485........570

Florida sales tax will be added to short course fees.


How to Contact SPIE
-------------------
If you have further questions, or need assistance, please send a
message to info-optolink-service at mom.spie.org. You will receive a
response from an SPIE staff member.


Join SPIE Today
---------------
Keep in touch with the dynamic world of optics and optoelectronics
by becoming a member of SPIE.

Full SPIE Membership
Joining SPIE as a full member provides you with many benefits,
including:

* Voting privileges
* Eligibility to hold SPIE office.
* Subscription to OE Reports
* Subscription to Optical Engineering, SPIE's monthly journal
* Annual SPIE Member Guide
* Full member discounts (~20%) on SPIE publications
* Full member discounts (~15%) on SPIE conferences and short
  courses
* Discounts on publications from other publishers as available
* Member rates for SPIE-cosponsored technical events

$85 in North America/$95 outside North America (Student Memberships
and Associate Student Memberships available at reduced rates.)


Working Group Membership
------------------------
Working Groups are interactive networks that foster professional
contacts and information flow among technically related
individuals, groups, companies, and institutions in specific areas
of technology.

Individual Membership ($15) Group Memberships and Corporate
Memberships are available. Contact SPIE for complete list of
working groups and benefits.

-------------------------------------------------------------------

              Abstract Due Date:  19 September 1994
    On-Site Proceedings Manuscript Due Date:  23 January 1995

       Manuscript due date for on-site proceedings must be
                      strictly observed.

------------------------------------------------------------------
For a complete text of the Announcement and Call for Papers for
SPIE's International Symposium on Aerospace/Defense Sensing and
Dual-Use Photonics, contact SPIE at either the European Office, or
International Headquarters addresses below.

Contact addresses:
SPIE in Europe:
  SPIE European Office
  c/o HIB-INFONET
  P.O. Box 4463
  N-5028 Bergen, Norway
  Phone: 47 55 54 37 84
  Fax: 47 55 96 21 75
  E-mail: spie at hibinc.no

SPIE International Headquarters
  P.O. Box 10
  Bellingham, WA  98227-0010 USA
  Phone: 206/676-3290
  Fax:  206/647-1445
  E-mail: spie at spie.org
  Telnet/FTP: spie.org
  World Wide Web URL: http://www.spie.org

-------------------------------------------------------------------
SPIE--The International Society for Optical Engineering

SPIE is a nonprofit society dedicated to advancing engineering and
scientific applications of optical, electro-optical, and
optoelectronic instrumentation, systems, and technology. Its
members are scientists, engineers, and users interested in the
reduction to practice of these technologies. SPIE provides the
means for communicating new developments and applications to the
scientific, engineering, and user communities through its
publications, symposia, and short courses.

SPIE is dedicated to bringing you quality electronic media and
online services.
-------------------------------------------------------------------

Dennis W. Ruck               Air Force Institute of Technology
d.ruck at ieee.org              Wright-Patterson AFB, Ohio
AFIT/ENG, Bldg 642, 2950 P ST, Wright-Patterson AFB OH 45433-7765
Ph. (513) 255-6565 ext. 4285   Fax (513) 476-4055


From UBJTP69 at CCS.BBK.AC.UK  Thu Aug 11 14:41:00 1994
From: UBJTP69 at CCS.BBK.AC.UK (Gareth)
Date: Thu, 11 Aug 94 14:41 BST
Subject: Thesis printing problem
Message-ID: <mailman.698.1149591314.29955.connectionists@cs.cmu.edu>

It has been reported that the file gaskell.thesis.ps.Z in the neuroprose
archive will not print out from some unix systems.  If this is the case,
edit the file and delete the initial character (^D).  The file should
then print out properly.

Sorry for any inconvenience.

Gareth Gaskell

From pluto at cs.ucsd.edu  Thu Aug 11 16:49:04 1994
From: pluto at cs.ucsd.edu (Mark Plutowski)
Date: Thu, 11 Aug 94 13:49:04 -0700
Subject: Paper in Neuroprose: estimating generalization with cross-validation
Message-ID: <9408112049.AA14173@beowulf>

FTP-host: archive.cis.ohio-state.edu
FTP-filename: /pub/neuroprose/pluto.imse.ps.Z

|This is related to the generalization debate on Machine Learning List.|


By popular request, the following paper has been placed in Neuroprose.

Title: 	   "Cross-validation estimates integrated mean squared error." 

Authors:   Plutowski , M. (1,3), S. Sakata (2), H. White (2,3). 
	   (1) Computer Science and Engineering
	   (2) Economics
	   (3) Institute for Neural Computation
	   (All at UCSD).


A discussion on the Machine Learning List prompted the question 

"Have theoretical conditions been established under which 
cross-validation is justified?"  

The answer is "Yes."  The statistical literature abounds with
application-specific and model-specific demonstrations that
cross-validation is statistically accurate and precise for use as 
a real-world estimate of an ideal measure of generalization known
as Integrated Mean Squared Error (IMSE).  IMSE is the
average mean squared error, averaged over all training sets of 
a particular size.   IMSE is closely related to Prediction Risk,
(aka statistical risk) therefore such results are applicable to 
statistical risk as well (as averaged over training sets of a
particular size).  See Plutowski's thesis in Neuroprose/Thesis for
explicit relationship between IMSE and statistical risk.

This paper extends such results to apply to nonlinear regression 
in general.  Strong convergence (w.p.1) and unbiasedness are proved.

The key assumption is that the training and test data be
independent and identically distributed (i.i.d.) - therefore, 
data must be drawn from the same space (stationarity), 
independent of previously sampled datum.

Note that if training data are explicitly excluded from the
test sample, then the i.i.d. assumption does not hold,
since in this case the test sample is drawn from a space that 
depends upon (is conditioned by) particular choice of training sample.  

Therefore, the measure of generalization referred to in the raging 
debate on the Machine Learning List would not meet the conditions 
employed by the results in this paper.


Filename:  pluto.imse.ps.Z.  
Title: 	   "Cross-validation estimates integrated mean squared error." 
File size: 97K compressed, 242K uncompressed.  
	   17 single-spaced pages (8 pages of text, the remainder is a
	   mathematical appendix).

Email contact: pluto at cs.ucsd.edu.  

SUBJECT: theorems proving cross-validation is a statistically 
accurate and precise estimator of an ideal measure of generalization.
Abridged version of this appeared in NIPS 6.


= Mark Plutowski


PS:  Sorry, no hard copies available.


From giles at research.nj.nec.com  Thu Aug 11 19:01:15 1994
From: giles at research.nj.nec.com (Lee Giles)
Date: Thu, 11 Aug 94 19:01:15 EDT
Subject: Corrected ps file of previously announced TR.
Message-ID: <9408112301.AA10157@fuzzy>


It seems that a figure in the postscript file of this TR 
generated printing problems for certain printers. We changed the figure 
to eliminate this problem. We apologize to anyone who was inconvenienced. 
In the revised TR, we also included some references that initially were 
unintentionally excluded.

Lee Giles, Bill Horne, Tsungnan Lin


_________________________________________________________________________________________



                 Learning a Class of Large Finite State Machines
                      with a Recurrent Neural Network

        UNIVERSITY OF MARYLAND TECHNICAL REPORT UMIACS-TR-94-94 AND CS-TR-3328

                 C. L. Giles[1,2], B. G. Horne[1], T. Lin[1,3] 

      [1] NEC Research Institute, 4 Independence Way, Princeton, NJ  08540
            [2]UMIACS, University of Maryland, College Park, MD 20742
          [3] EE Department, Princeton University, Princeton, NJ 08540

                      {giles,horne,lin}@research.nj.nec.com
                
One of the issues in any learning model is how it scales with problem size. Neural 
networks have not been immune to scaling issues. We show that a dynamically-
driven discrete-time recurrent network (DRNN) can learn rather large grammatical 
inference problems when the strings of a finite memory machine (FMM) are 
encoded as temporal sequences. FMMs are a subclass of finite state machines 
which have a finite memory or a finite order of inputs and outputs. The DRNN that 
learns the FMM is a neural network that maps directly from the sequential 
machine implementation of the FMM. It has feedback only from the output and not 
from any hidden units; an example is the recurrent network of Narendra and 
Parthasarathy. (FMMs that have zero order in the feedback of outputs are called 
definite memory machines and are analogous to Time-delay or Finite Impulse 
Response neural networks.) Due to their topology these DRNNs are as least as 
powerful as any sequential machine implementation of a FMM and should be 
capable of representing any FMM. We choose to learn ``particular FMMs.\' Specif
ically, these FMMs have a large number of states (simulations are for $256$ and 
$512$ state FMMs) but have minimal order, relatively small depth and little logic 
when the FMM is implemented as a sequential machine. Simulations for the num
ber of training examples versus generalization performance and FMM extraction 
size show that the number of training samples necessary for perfect generalization 
is less than that sufficient to completely characterize the FMM to be learned. This 
is in a sense a best case learning problem since any arbitrarily chosen FMM with a 
minimal number of states would have much more order and string depth and most 
likely require more logic in its sequential machine implementation.


--------------------------------------------------------------------------------------

                          FTP INSTRUCTIONS

                unix> ftp cs.umd.edu (128.8.128.8)
                Name: anonymous
                Password: (your_userid at your_site)
                ftp> cd pub/papers/TRs
                ftp> binary
                ftp> get 3328.ps.Z
                ftp> quit
                unix> uncompress 3328.ps.Z

                                 OR

                unix> ftp external.nj.nec.com (138.15.10.100)
                Name: anonymous
                Password: (your_userid at your_site)
                ftp> cd pub/giles/papers
                ftp> binary
                ftp> get large.fsm.ps.Z
                ftp> quit
                unix> uncompress large.fsm.ps.Z

--------------------------------------------------------------------------------------


--                                 
C. Lee Giles / NEC Research Institute / 4 Independence Way
Princeton, NJ 08540 / 609-951-2642 / Fax 2482
==




From grino at ic.upc.es  Fri Aug 12 16:11:50 1994
From: grino at ic.upc.es (Robert Grino)
Date: Fri, 12 Aug 1994 16:11:50 UTC+0100
Subject: Paper in Neuroprose: Nonlinear System Identification Using Additive
 Dynamic Neural Networks
Message-ID: <232*/S=grino/OU=ic/O=upc/PRMD=iris/ADMD=mensatex/C=es/@MHS>

FTP-host: archive.cis.ohio-state.edu
FTP-filename: /pub/neuroprose/grino.sysid.ps.Z

The file grino.sysid.ps.Z is now available for
copying from the Neuroprose repository:

NONLINEAR SYSTEM IDENTIFICATION USING ADDITIVE DYNAMIC NEURAL NETWORKS
R. Grino (grino at ic.upc.es)
Instituto de Cibernetica - ESAII
Univ. Politecnica Catalunya
Diagonal, 647, 2nd floor
08028-Barcelona, SPAIN
(Reprint of a SICICA'94 (IFAC) paper)


ABSTRACT:
In this work additive dynamic neural models are used for the 
identification of nonlinear plants in on-line operation. In order to accomplish
this task a gradient parameter adaptation method based in sensitivity 
analysis is formulated taking into account that the parameters of the model 
are arranged in matrix form. This methodology is applied to several nonlinear 
systems in simulation and with a real dataset to verify its performance.

=============================================================================
   Robert Grino                   E-mail: grino at ic.upc.es
   Instituto de Cibernetica               
   Diagonal,647, 2nd floor        FAX number: (343) 4016605
   08028 - Barcelona              
       SPAIN    
=============================================================================

From dhw at santafe.edu  Fri Aug 12 16:32:37 1994
From: dhw at santafe.edu (dhw@santafe.edu)
Date: Fri, 12 Aug 94 14:32:37 MDT
Subject: No subject
Message-ID: <9408122032.AA06773@chimayo.santafe.edu>

Mark Plutowski recently said on connectionist:


>>>
A discussion on the Machine Learning List prompted the question 

"Have theoretical conditions been established under which 
cross-validation is justified?"  

The answer is "Yes." 
>>>



Mark is being polite by not using names; I am the one he is
(implicitly) taking to task, for the following comment on the
ML list:

>>>
. an assumption *must always* be
present if we are to have any belief in learnability in the problem
at hand...
However, to give just one example, nobody has yet delineated
just what those assumptions are for the technique of cross-validation.
>>>

Mark is completely correct in his (implicit) criticism.

As he says, there has in fact been decades of work analyzing 
cross-validation from a sampling theory perspective. Mark's thesis is 
a major contribution to this literature. Any implications coming from my
message that such literature doesn't exist or is somehow invalid
are completely mistaken and were not intended. (Indeed, I've had 
several very illuminating discussions with Mark about his thesis!)

The only defense for the imprecision of my comment is that I made 
it in the context of the ongoing discussion on the ML list that 
Mark referred to. That discussion concerned off-training set error 
rather than iid error, so my comments implicitly assumed off-training 
set error. And as Mark notes, his results (and the others in the 
literature) don't extend to that kind of error.

(Another important distinction between the framework Mark uses and the
implicit framework in the ML list discussion is that the latter has
been concerned w/ zero-one loss, whereas Mark's work concentrates on
quadratic loss. The no-free-lunch results being discussed in the ML
list change form drastically if one uses quadratic rather than
zero-one loss. That should be no surprise, given, for example, the
results in Michael Perrone's work.)

While on the subject of cross-validation and iid error though, it's
interesting to note that there is still much to be understood. For
example, in the average-data scenario of sampling theory statistics
that Mark uses, asymptotic properties are better understood than
finite data properties. And in the Bayesian this-data framework, very
little is known for any data-size regime (though some work by Dawid on
this subject comes to mind).




David Wolpert




From mccauley at ecn.purdue.edu  Thu Aug 11 21:51:21 1994
From: mccauley at ecn.purdue.edu (James Darrell McCauley)
Date: Thu, 11 Aug 1994 20:51:21 -0500
Subject: connectionists suggestion
In-Reply-To: <9408112049.AA14173@beowulf>
Message-ID: <199408120151.UAA03022@alx3.ecn.purdue.edu>

Mark Plutowski (pluto at cs.ucsd.edu) writes on 11 Aug 94:
>FTP-host: archive.cis.ohio-state.edu
>FTP-filename: /pub/neuroprose/pluto.imse.ps.Z

May I suggest that posters be encouraged to use
the URL format and *not* post instructions on the
usage of ftp? I.e., the following would be sufficient:

ftp://archive.cis.ohio-state.edu/pub/neuroprose/pluto.imse.ps.Z

(this way readers may just cut/paste the string into xmosaic
or whatever browser they use).

Not picking on Dr Plutoski, just a suggestion that came to mind...

--Darrell McCauley, aka jdm5548 at diamond.tamu.edu

James Darrell McCauley, Purdue Univ, West Lafayette, IN 47907-1146, USA
mccauley at ecn.purdue.edu, mccauley%ecn at purccvm.bitnet, pur-ee!mccauley

From harnad at Princeton.EDU  Fri Aug 12 23:53:54 1994
From: harnad at Princeton.EDU (Stevan Harnad)
Date: Fri, 12 Aug 94 23:53:54 EDT
Subject: French Cognitive Science Conference
Message-ID: <9408130353.AA25579@clarity.Princeton.EDU>

From: payette at uranus.atoci.uqam.ca
Subject: French International Cognitive Science Conference

                              ANNOUNCEMENT
        The Seventh Colloquium of the Jacques Cartier Center
                              Lyon, France. 

            THE COGNITIVE SCIENCES: FROM COMPUTATIONAL MODELS  
                       TO THE  PHILOSOPHY OF MIND                        
               
                           under the aegis of:
            the Pole Rhones-Alpes of the Cognitive Sciences, 
      Programme Interdisciplinaire de Recherche Cognisciences,CNRS
                    Universite du Quebec a Montreal 
                        Universite de Montreal 
                       Universite Joseph Fourier
                       Universite Claude Bernard

Scientific committee:
Denis Fisette  (Universite du Quebec a Montreal, Quebec)
Marc Jeannerod (Universite Claude Bernard, Lyon)
Daniel Laurier (Universite de Montreal, Quebec)
Daniel Payette (Universite du Quebec a Montreal, Quebec)
Vincent Rialle (Universite Joseph Fourier, Grenoble)
Guy Tiberghien (Universite Pierre Mendes-France, Grenoble)

Coordination in North America: Daniel Payette and Denis Fisette  
Universite du Quebec a Montreal, Dpt de Philosophie, Dpt Psychology; C.P.
8888,Succ A, Montreal (Quebec) H3C-3P8, Canada; 
E.mail : payette at uranus.atoci.uqam.ca; 
tel (+514) 987 8418;  Fax: (+514) 9876721 

Coordination in Europe: Vincent Rialle
Universite J. Fourier, Labo.TIMC-IMAG, Faculte de Medecine, 38706 LaTronche
Cedex  E.mail: Vincent.Rialle at imag.fr; 
Tel. (+33) 76 63 71 87;  Fax. (+33) 76 51 8667

DATES: Wednesday, November 30th to Friday, December 2nd 1994

CONFERENCE SITE:  
Amphitheatre CHARLES BERAUDIER. Conseil Regional RHONE-ALPES,78 route de
Paris 69751 CHARBONNIERES-les-BAINS. France

*Talks will only be given by invited speakers. (Simultaneous
French-English and English-French will be provided).
                         
                THEME OF COLLOQUIUM

The modeling of mental processes in the various human cognitive
activities has generated increasing interest in the scientific world
today. Cognitive models, cognitive simulations, auto-organization,
adaptation, emergence, genetic selection, Darwinian mentalism and
enaction are active research topics in neurological and psychological
theory.

The cognitive sciences offer a continuum of research extending from the
engineering sciences to the philosophy of mind, including the
neurosciences, cognitive psychology, linguistics, semantics, semiotics
and artificial intelligence. Three subconferences will organize
themselves around the following major complementary themes: (i)
Modeling (cognitive and brain functions), (ii) Philosophy of Mind and
Epistemology, and (iii) Applications (AI, technical and computational
engineering).

(i) Modeling is a point of intersection for all these specialties
because it includes the modeling of functions and dysfunctions of the
central nervous system, the neurocomputer sciences, the modeling of
psychocognitive and mental processes, the emergence of intentional
structure on the basis of biological structure, enaction, genetic
algorithms, neural networks, artificial "life," etc.

(ii) The philosophical and epistemological subcomponent poses questions
like the following:  Can we elaborate mathematical models of the mind
and use them to describe and explain human behavior? Are we aiming
toward a mathematical model of the mind? Can we capture the formal
principles of the development and emergence of cognition? Can we
technologically recreate thought? Is the computational symbolic
paradigm, which has imposed itself for the last decades, still a
powerful conceptual tool or is it proving too reductionistic and if so,
how? What is the epistemological status of, for example, the
alternative proposed by the parallel distributed model to the
computational models of classical cognitivism? What is the relation os
the modeling activity of the cognitive and neurosciences and human
experience?

(iii) The applications subconference will consider practical domains in
which scientific results have been applied in the treatment of
language, the automated cognitive analyses of textual documents (an
intersection of linguistics, semantics, semiotics and artificial
intelligence), aids to decision making, applications in sensory
information processing, etc.

PREPROGRAM

WEDNESDAY, 30 November 1994
8h15 - 8h30
Allocution d'accueil du Conseil Regional
8h30 - 9h
Guy Tiberghien (Universite Pierre Mendes-France, Grenoble)
Introduction

SESSION 1 : Modelisation neuro et psycho-cognitives

9h - 9h-30
Jean Francois Le Ny (Universite Paris-Sud, psychologie cognitive)
Pourquoi les modeles cognitifs devraient-ils etre calculatoires ?
9h30 - 9h45 Discussion
9h-45 - 10h15
Marc Jeannerod (Universite Claude Bernard, Lyon, neurosciences)
Le cerveau representationnel
10h15 - 10h30 Discussion
10h30 - 10h45 PAUSE
10h45 - 11h15
Zenon Pylyshyn (Rutgers University, USA, psychologie cognitive)
What's in the Mind? A Computational Approach to a Ancient Question.
11h15 - 11h30 Discussion
11h30 - 12h00
Stevan Harnad (Princeton University, psychologie cognitive)
Modeles, mobiles et mentalite
12h00 - 12h15 Discussion

MEAL

14h00 - 14h30
Michel Imbert (Universite Paul Sabatier, Toulouse, neurosciences)
De l'etude du cerveau a la comprehension de l'esprit
14h30 - 14h45 Discussion
14h45 - 15h15
Guy Tiberghien (Univers Pierre Mendes-France,Grenoble,psychologie
cognitive)
Connexionnisme: stade supreme du behaviorisme ?
15h15 - 15h30 Discussion
15h30 - 15h45 PAUSE
15h45 - 16h15
Jacques Demongeot  (Universite Joseph Fourier, Grenoble, neurosciences)
Memoire d'evocation dans les reseaux de neurones
16h15 - 16h30 Discussion
16h30 - 17h00
Bennet Murdock (Universite de Toronto, psychologie cognitive)
THE ROLE OF FORMAL MODELS IN MEMORY RESEARCH
17h00 - 17h15 Discussion
17h15 - 17h45
Robert Proulx (Universite du Quebec a Montreal, neuro-psychologie)
Plausibilite biologique de certains systemes de categorisation adaptative a
base de reseaux de neurones
17h45 - 18h00 Discussion

TUESDAY, December 1
Session 2 : Epistemology, Philosophy of Mind and Cognition

9h - 9h30
Elisabeth Pacherie (Universite de Provence, CNRS & CREA, Paris)
Domaines cognitifs et modularite
9h30 - 9h45 Discussion
9h-45 - 10h15
Pierre Livet (Universite de Provence & CREA, Paris, philosophie)
Categorisation et connexionnisme
10h15 - 10h30 Discussion
10h30 - 10h45 PAUSE
Normand Lacharite (Universite du Quebec a Montreal, epistemologie)
10h45 - 11h15
Conflits de modeles en theorie de la representation
11h15 - 11h30 Discussion
11h30 - 12h00
Peter Gardenfors (Lund University, Suede, philosophie)
Language and the Evolution of Mind
12h15 - 12h15 Discussion

MEAL

14h00 - 14h30
Andy Clark (Washington University, philosophie)
Wild Cognition: Putting Representation in its Place
14h30 - 14h45 Discussion
14h45 - 15h15
Kevin Mulligan (Universite de Geneve, Suisse, philosophie)
Constance perceptuelle et contenu spatial
15h15 - 15h30 Discussion
15h30 - 15h45 PAUSE
15h45 - 16h15
Ronald De Sousa (Universite de Toronto, epistemologie)
La rationalite: un concept normatif ou descriptif ?
16h15 - 16h30 Discussion
16h30 - 17h00
Daniel Laurier (Universite de Montreal, philosophie)
Rationalite et naturalisme
17h00 - 17h15 Discussion
17h15 - 17h45
Joelle Proust (CNRS & CREA, Paris, philosophie)
Un modele naturaliste de l'intentionnalite
17h45 - 18h00 Discussion

FRIDAY, December 2
Session 3: Modelisation IA, Traitement du langage, et semantique cognitive

Paul Jorion (Maison des Sciences de l'Homme, Paris, psychologie cognitive)
9h - 9h30
Modelisation du reseau mnesique : une utilisation minimaliste de l'IA
9h30 - 9h45 Discussion
9h-45 - 10h15
Bernard Amy (Universite Joseph Fourier, Grenoble, connexionnisme)
La place des reseaux neuronaux dans l'IA
10h15 - 10h30 Discussion
10h30 - 10h45 PAUSE 
10h45 - 11h15
Paul Bourgine (CEMAGREF, Paris-Antony, IA-modelisation)
Co-evolution et emergence du soi
11h15 - 11h30 Discussion
11h30 - 12h00
Paul Pietroski (Universite McGill, Canada, philosophie)
What can linguistics teach us about belief
12h00 - 12h15 Discussion
MEAL
14h00 - 14h30
Le paradigme hermeneutique et la mediation semiotique
Francois Rastier (Institut National de la Langue Francaise, CNRS,
                  linguistique computationnelle)
14h30 - 14h45 Discussion
14h45 - 15h15
L'impact des perspectives cognitives dans le traitement de l'information
Jean-Guy Meunier (Universite du Quebec a Montreal, semiotique)
15h15 - 15h30 Discussion
15h30 - 15h45 PAUSE
15h45 -  16h15
Guy Denhiere (Universite Paris VIII, psychologie cognitive)
Isabelle Tapiero (Universite Lyon II, psychologie cognitive)
La signification comme structure emergente : de l'acces au lexique a la
comprehension de textes
16h15 - 16h30 Discussion
16h30 - 17h00
Paul Freedman (Centre de Recherche en Informatique de Montreal, IA)
La vision artificielle: le traitement intelligent de documents
17h00 - 17h15 Discussion
17h15 - 17h45
Denis Vernant (Universite Pierre Mendes-France, Grenoble, philosophie)
L'intelligence de la machine et sa capacite dialogique
17h45 - 18h00 Discussion

18h00: END OF COLLOQUIUM

------------------------------------------------------------------------
-ADMISSION FEES-
(Includes:access to the conference room, meals and the colloquium
documents)

Individuals-------------------------------------------------1500FF
Student (join proof of eligibility with registration)------- 500FF

----------------------------------------------------------------------------
--

                           REGISTRATION BULLETIN
(The Cognitive Sciences:From computational models to philosophy of mind)

Name:___________________________________________________________________
             
Status:_____________________________________                              
Institution/Company_________________________                               

Complete Address_________________________________________________________
Fax:________________________                                            
Phone :______________________                                     
@mail number__________________________________

Enclosed : 	 Check	 or money order of (_____________________FF)
(Make check or money order payable to CENTRE JACQUES CARTIER)

-Send information on possibilities of housing in Lyon(______)
_Send me the colloquium brochure (_____)

-November 30 meal __
-December 1, meal __
-December 2, meal __

RETURN TO: CENTRE JACQUES CARTIER, 86 rue Pasteur, 69365 Lyon Cedex 07,  
                                    France. Phone:(33) 78 69 72 21


From dhw at santafe.edu  Sat Aug 13 01:19:30 1994
From: dhw at santafe.edu (dhw@santafe.edu)
Date: Fri, 12 Aug 94 23:19:30 MDT
Subject: Cross-validation theory
Message-ID: <9408130519.AA08172@chimayo.santafe.edu>



Mark Plutowski recently said on connectionist:


>>>
A discussion on the Machine Learning List prompted the question

"Have theoretical conditions been established under which
cross-validation is justified?"

The answer is "Yes."
>>>

As Mark goes on to point out, there are decades
of work on cross-validation from a likelihood-driven
sampling theory perspective. Indeed, Mark's thesis is a
major addition to that literature.

Mark then correctly notes that this literature doesn't *directly* 
apply to the discusson on the ML list, since that discussion involves
off-training set rather than iid error.

It should be noted that there is another important distinction between 
the framework Mark uses and the implicit framework in the ML list 
discussion; the latter has been concerned w/ zero-one loss, whereas 
Mark's work concentrates on quadratic loss. The no-free-lunch results 
being discussed in the ML list change form drastically if one uses 
quadratic rather than zero-one loss. That should not be too surprising, 
given, for example, the results in Michael Perrone's work involving
quadratic loss.

It's also worth noting that even in the regime of iid error and quadratic
loss, there is still much to be understood. For example, in the 
average-data scenario of sampling theory statistics that Mark uses, 
asymptotic properties are better understood than finite data properties. 
And in the Bayesian this-data framework, very little is known for any 
data-size regime (though some work by Dawid on this subject comes to mind).




David Wolpert

From mdg at magi.ncsl.nist.gov  Tue Aug 16 11:58:28 1994
From: mdg at magi.ncsl.nist.gov (Mike Garris x2928)
Date: Tue, 16 Aug 94 11:58:28 EDT
Subject: Announcement: Public Domain OCR
Message-ID: <9408161558.AA02660@magi.ncsl.nist.gov>

                   ANNOUNCEMENT - PUBLIC DOMAIN OCR

             NIST FORM-BASED HANDPRINT RECOGNITION SYSTEM
             ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

              Michael D. Garris (mdg at magi.ncsl.nist.gov)
     James L. Blue, Gerald T. Candela, Darrin L. Dimmick, Jon Geist,
       Patrick J. Grother, Stanley A. Janet, and Charles L. Wilson

            National Institute of Standards and Technology,
                        Building 225, Room A216
                      Gaithersburg, Maryland 20899
            Phone: (301)975-2928          FAX: (301)840-1357


The National Institute of Standards and Technology (NIST) has developed a
standard reference form-based handprint recognition system for evaluating
optical character recognition (OCR). NIST is making this recognition system
freely available to the general public on an ISO-9660 format CD-ROM. The
recognition system processes the Handwriting Sample Forms distributed with
NIST Special Database 1 and NIST Special Database 3. The system reads
handprinted fields containing digits, lower case letters, upper case letters,
and reads a text paragraph containing the Preamble to the U.S. Constitution.

This is a source code distribution written primarily in C and is organized
into 11 libraries. There are approximately 19,000 lines of code supporting
more than 550 subroutines. Source code is provided for form registration,
form removal, field isolation, field segmentation, character normalization,
feature extraction, character classification, and dictionary-based post-
processing. A host of data structures and low-level utilities are also
provided. These utilities include the application of CCITT Group 4 decompres-
sion, IHead file manipulation, spatial histograms, Least-Squares fitting,
spatial zooming, connected components, Karhunen Loeve (KL) feature extraction,
optimized Probabilistic Neural Network classification, multiple-key sorting,
Levenstein distance dynamic string alignment, and dictionary-based post-
processing. Two supporting programs are provided that compute eigenvectors
and KL feature vectors for training classifiers. Unlike the recognition
system (which is written entirely in C), these two programs contain FORTRAN
subroutines. To support these programs, a training set of 168,365 segmented
and labeled character images is provided. About 1000 writers contributed to
this training set.

The NIST standard reference recognition system is designed to run on UNIX
workstations and has been successfully compiled and tested on a Digital
Equipment Corporation (DEC) Alpha, Hewlett Packard (HP) Model 712/80, IBM
RS6000, Silicon Graphics Incorporated (SGI) Indigo 2, SGI Onyx, SGI Challenge,
Sun Microsystems (Sun) IPC, Sun SPARCstation 2, Sun 4/470, and a Sun SPARC-
station 10.** Scripts for installation and compilation on these architectures
are provided with this distribution.

A CD-ROM distribution of this standard reference system can be obtained free
of charge by sending a letter of request to Michael D. Garris at the address
above. The letter, preferably on company letterhead, should identify the
requesting organization or individuals. This system or any portion of this
system may be used without restrictions. However, redistribution of this
standard reference recognition system is strongly discouraged as any
subsequent corrections or updates will be sent to registered recipients only.
This software was produced by NIST, an agency of the U.S. government, and by
statute is not subject to copyright in the United States. Recipients of this
software assume all responsibilities associated with its operation,
modification, and maintenance.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
** Specific hardware and software products identified were used in order to
   adequately support the development of this technology. In no case does
   such identification imply recommendation or endorsement by the National
   Institute of Standards and Technology, nor does it imply that the
   equipment identified is necessarily the best available for the purpose.

From terry at salk.edu  Tue Aug 16 20:12:13 1994
From: terry at salk.edu (Terry Sejnowski)
Date: Tue, 16 Aug 94 17:12:13 PDT
Subject: Neural Computation 6:5
Message-ID: <9408170012.AA07668@salk.edu>

NEURAL COMPUTATION
September 1994 Volume 6 Number 5

Articles:

A Bayesian Analysis of Self-Organizing Maps
Stephen P. Luttrell

Network Amplification of Local Fluctuations Causes High Spike Rate
Variability, Fractal Firing Patterns and Oscillatory Local Field Potentials
Marius Usher, Martin Stemmler, Christof Koch and Zeev Olami

Note:

Statistical Analysis of an Autoassociative Memory Network
A. M. N. Fu

Letters:

Loading Deep Networks is Hard
Jiri Sima

Measuring the VC-dimension of a Learning Machine
Vladimir Vapnik, Esther Levin and Yann Le Cun

Neural Nets with Superlinear VC-Dimension
Wolfgang Maass

A Novel Design Method for Multilayer Feedforward Neural Networks
Jihong Lee

An Internal Mechanism for Detecting Parasite Attractors in a Hopfield Network
Jean-Dominique Gascuel, Bahram Moobed and Michel Weinfeld

On Langevin Updating in Multilayer Perceptrons
Thorsteinn Rognvaldsson

Probabilistic Winner-Take-All Learning Algorithm for Radial-Basis-Function
Neural Classifiers
Hossam Osman and Moustafa M. Fahmy

Realization of the "Weak Rod" by a Double Layer Parallel Network
T. Matsumoto and K. Kondo

Learning in Neural Networks with Material synapses
Daniel J. Amit and Stefano Fusi

Model Based on Extracellular Potassium for Spontaneous Synchronous Activity
in Developing Retinas
Pierre-Yves Burgi and Norberto M. Grzywacz

Bayesian Modeling and Classification of Neural Signals
Michael S. Lewicki

-----
 
SUBSCRIPTIONS - 1994 - VOLUME 6 - BIMONTHLY (6 issues)
 
______ $40     Student and Retired
______ $65     Individual
______ $166    Institution
 
Add $22 for postage and handling outside USA (+7% GST for Canada).
 
(Back issues from Volumes 1-5 are regularly available for $28 each
to institutions and $14 each for individuals
Add $5 for postage per issue outside USA (+7% GST for Canada)
 
MIT Press Journals, 55 Hayward Street, Cambridge, MA 02142.
Tel: (617) 253-2889  FAX: (617) 258-6779  e-mail: hiscox at mitvma.mit.edu
 
-----

From ucganlb at ucl.ac.uk  Wed Aug 17 05:01:47 1994
From: ucganlb at ucl.ac.uk (Dr Neil Burgess - Anatomy UCL London)
Date: Wed, 17 Aug 94 10:01:47 +0100
Subject: pre-print in neuroprose: hippocampus - spatial models
Message-ID: <93386.9408170901@link-1.ts.bcc.ac.uk>


ftp://archive.cis.ohio-state.edu/pub/neuroprose/burgess.hbtnn.ps.Z

The above/below file has been put on neuroprose for anonymous ftp, www
or whatever, contact n.burgess at ucl.ac.uk with any retrieval problems.
All the best,
Neil

		HIPPOCAMPUS - SPATIAL MODELS
	Neil Burgess, Michael Recce & John O'Keefe
	Dept. of Anatomy, University College London,
		   London WC1E 6BT, U.K.
		e-mail: n.burgess at ucl.ac.uk

This is a brief review of models of the hippocampus, focussing on 
spatial aspects of cell-firing and hippocampal function. 
To appear in `The Handbook of Brain Theory and Neural Networks'
(M. A. Arbib Ed.), Bradford Books/ MIT Press, 1995, and restricted in 
length and number of citations accordingly.

8 pages, 0.46 Mbytes uncompressed, 
hard-copies availible in extreme circumstances only.


From uzimmer at informatik.uni-kl.de  Wed Aug 17 12:14:01 1994
From: uzimmer at informatik.uni-kl.de (Uwe R. Zimmer, AG vP)
Date: Wed, 17 Aug 94 17:14:01 +0100
Subject: Papers available (world modelling, mobile robots)
Message-ID: <940817.171401.723@ag_vp_file_server.informatik.uni-kl.de>

A couple of actual papers about:

--------------------------------------------------------------
---     Learning, Robotics, Visual Search, Navigation,     ---
---        Topologic Maps & Robust Mobile Robots           ---
---                    Neural Networks                     ---
--------------------------------------------------------------

are now available via FTP:

------------------------------------------------------------------------
--- Comparing World-Modelling Strategies for Autonomous Mobile Robots
------------------------------------------------------------------------

---   File name is : Zimmer.Comparison.ps.Z   ---

         IWK `94, Ilmenau, Germany, September 27 - 30, 1994

   Comparing World-Modelling Strategies for Autonomous Mobile Robots

                Uwe R. Zimmer & Ewald von Puttkamer

The focus of this paper is on strategies for adapting a couple of
internal representations to the actual environment of a mobile robot.

From svc at demos.lanl.gov  Tue Aug 16 15:22:10 1994
From: svc at demos.lanl.gov (Stephen Coggeshall)
Date: Tue, 16 Aug 94 13:22:10 MDT
Subject: graduate student positions
Message-ID: <9408161922.AA13573@demos.lanl.gov>

At Los Alamos National Laboratory in New Mexico we have a small research
effort using adaptive computational models for a variety of projects.  At
this time there may be the possibility of a few graduate student positions
available specifically for pattern recognition/data base mining with
application to financial problems. We are interested in highly motivated,
self-directed students with strong backgrounds in programming, math, neural
net applications.  Interested parties can contact Steve (svc at lanl.gov).
Please describe briefly your past work and current interests, as well as
your availability.

From R.Gaizauskas at dcs.shef.ac.uk  Fri Aug 19 17:12:12 1994
From: R.Gaizauskas at dcs.shef.ac.uk (Robert John Gaizauskas)
Date: Fri, 19 Aug 94 17:12:12 BST
Subject: AISB-95 WORKSHOP/TUTORIAL CALL
Message-ID: <9408191612.AA03843@dcs.shef.ac.uk>


        -------------------------------------------------
        AISB-95: CALL FOR WORKSHOP AND TUTORIAL PROPOSALS
        -------------------------------------------------

             Call for Workshop Proposals: AISB-95

                 University of Sheffield,
                    Sheffield, England

                    April 3 -- 4, 1995

                   Society for the Study of
      Artificial Intelligence and Simulation of Behaviour (SSAISB)


The AISB Committee invites proposals for workshops to be held in conjunction
with the Tenth Biennial Conference on AI and Cognitive Science (AISB-95).
While the main conference will run for three days from Wednesday, April 5 to
Friday, April 7, the workshops will be held on the two days preceding the main
event: Monday, April 3 and Tuesday, April 4.

The main conference has the theme "Hybrid Problems, Hybrid Solutions" (see the
main conference call) and while proposals for workshops related to that theme
would be particularly welcome, proposals are invited for workshops relating
to any aspect of Artificial Intelligence or the Simulation of Behaviour.

Proposals, from an individual or a pair of organisers, for workshops between
0.5 and 2 days long will be considered. Workshops will probably address topics
which are at the forefront of research, but not yet sufficiently developed to
warrant a full-scale conference.

 Submission:
 ----------
 A workshop proposal should contain the following information:

 1. Workshop Title

 2. A detailed outline of the workshop.
    This should include the necessary background and the potential
    target audience for the workshop and a justified estimate of the
    number of possible attendees. Please also state the length and
    preferred date(s) of the workshop. Specify any equipment requirements,
    indicating whether the organisers would be expected to meet them.

 3. A brief resume of the organiser(s).
    This should include: background in the research area,
    references to published work in the topic area and relevant
    experience, such as previous organisation or chairing of
    workshops.

 4. Administrative information.
    This should include: name, mailing address, phone number, fax, and
    email address if available.  In the case of multiple organisers,
    information for each organiser should be provided, but one organiser
    should be identified as the principal contact.

 5. A draft Call for Participation.
    This should serve the dual purposes of informing and attracting
    potential participants.


 The organisers of accepted workshops are responsible for issuing a call for
 participation, reviewing requests to participate and scheduling the workshop
 activities within the constraints set by the Workshop Organiser. They are
 also responsible for submitting a collated set of papers for their workshop
 to the Workshop Organiser.

 Dates:
 ------
 Intentions to organise a workshop should be made known to the
 Workshop Organiser as soon as possible.

 Proposals must be received by October 18th 1994.
 Decisions about topics and speakers will be made in early November.
 Collated sets of papers to be received by March 15th 1995.

 Proposals should be sent to:

    Dr. Robert Gaizauskas
    Department of Computer Science
    University of Sheffield
    211 Portobello Street
    Regent Court
    Sheffield S1 4DP
    U.K.

    email:  robertg at dcs.shef.ac.uk
    phone:  +44 (0)742 825572
    fax:    +44 (0)742 780972

Electronic submission (plain ascii text) is highly preferred, but
hard copy submission is also accepted, in which case 5 copies should
be submitted. Proposals should not exceed 2 sides of A4 (i.e. 120
lines of text approx.).

---------------------------------------------------------------------

             Call for Tutorial Proposals: AISB-95

                 University of Sheffield,
                    Sheffield, England

                    April 3 -- 4, 1995

                   Society for the Study of
      Artificial Intelligence and Simulation of Behaviour (SSAISB)


The AISB Committee invites proposals for Tutorials to be held in conjunction
with the Tenth Biennial Conference on AI and Cognitive Science (AISB-95).
While the main conference will run for three days from Wednesday, April 5 to
Friday, April 7, the tutorials will be held on the two days preceding the main
event: Monday, April 3 and Tuesday, April 4.

Proposals for full and half day tutorials, from an individual or pair of
presenters, will be considered. They may be offered both on standard topics
and on new and more advanced aspects of Artificial Intelligence or Simulation
of Behaviour. Anyone interested in presenting a tutorial should submit a
proposal to the Workshop Organiser Dr Robert Gaizauskas (addresses below).

 Submission:
 ----------
 A tutorial proposal should contain the following
 information:

 1. Tutorial Title

 2. A brief description of the tutorial, suitable for inclusion in
    a brochure.

 3. A detailed outline of the tutorial.
    This should include the necessary background and the potential
    target audience for the tutorial and a justified estimate of the
    number of possible attendees. Please also state the length and
    preferred date(s) of the tutorial. Specify any equipment requirements,
    indicating whether the organisers would be expected to meet them.

 4. A brief resume of the presenter(s).
    This should include: background in the tutorial area,
    references to published work in the topic area and relevant
    experience. Published work should, ideally, include a published
    tutorial-level article on the subject. Relevant experience
    is teaching experience, including previous conference tutorials
    or short courses presented.

 5. Administrative information.
    This should include: name, mailing address, phone number, fax, and
    email address if available.  In the case of multiple presenters,
    information for each presenter should be provided, but one presenter
    should be identified as the principal contact.

 The presenter(s) of accepted tutorials must submit a set of
 tutorial notes (which may include relevant tutorial-level
 publications) to the Workshop Organisers by March 15th 1995.

 Dates:
 ------
 Intentions to organise a tutorial should be made known to the
 the Workshop Organiser as soon as possible.

 Proposals must be received by October 18th 1994.
 Decisions about tutorial topics and speakers will be made in early November.
 Tutorial notes must be received by March 15th 1995.

 Proposals should be sent to:

    Dr. Robert Gaizauskas
    Department of Computer Science
    University of Sheffield
    211 Portobello Street
    Regent Court
    Sheffield S1 4DP
    U.K.

    email:  robertg at dcs.shef.ac.uk
    phone:  +44 (0)742 825572
    fax:    +44 (0)742 780972

Electronic submission (plain ascii text) is highly preferred, but
hard copy submission is also accepted, in which case 5 copies should
be submitted. Proposals should not exceed 2 sides of A4 (i.e. 120
lines of text approx.).

From bishopc at helios.aston.ac.uk  Fri Aug 19 09:33:57 1994
From: bishopc at helios.aston.ac.uk (bishopc)
Date: Fri, 19 Aug 1994 13:33:57 +0000
Subject: Research Associate - Software Support
Message-ID: <9192.9408191233@sun.aston.ac.uk>


-------------------------------------------------------------------


                     Aston University 

               Neural Computing Research Group


      RESEARCH ASSOCIATE - SYSTEMS AND SOFTWARE SUPPORT
      -------------------------------------------------


Applications are invited for a position as a Research Associate 
within the Neural Computing Research Group both to provide support for
the Group's local network of Sun workstations and to undertake 
software development and research in support of projects within
the Group. 

The Neural Computing Research Group currently comprises three
professors, two lecturers, three postdoctoral research fellows and 
ten postgraduate research students. In addition, two further 
Lecturerships have recently been advertised. Current research activity
focusses on principled approaches to neural computing, and spans 
a broad spectrum from theoretical foundations to industrial and
commercial applications. 

The ideal candidate will have significant experience of the UNIX
operating system and system maintenance, experience of software 
engineering in C++, and an understanding of neural networks.
The responsibilities of the successful candidate will be as 
follows:

(1) To provide system support for the Group's LAN of Sun UNIX
workstations and associated peripherals, ensurimg that an
efficient working environment is maintained.

(2) To support the development, testing and documentation of
the NetLib C++ library of neural network software.

(3) To assist with numerical experiments in support of research
projects, including industrial contracts. This aspect of the work 
may provide opportunities for joint publication in academic journals.

(4) To provide such other software support as may be required, 
such as the maintenance of LaTeX, provision of WWW pages for the
Group, etc.

Salaries will be 13,941 UK pounds or above, depending on
the experience and qualifications of the successful applicant.

If you wish to apply for this position, please send a CV, together 
with the names and addresses of 3 referees, to:

    Professor Chris Bishop
    Neural Computing Research Group
    Aston University
    Birmingham B4 7ET, U.K.
    Tel: 021 359 3611 ext. 4270
    Fax: 021 333 6215
    e-mail: c.m.bishop at aston.ac.uk

    closing date: 9 September 1994



From harnad at Princeton.EDU  Fri Aug 19 22:51:41 1994
From: harnad at Princeton.EDU (Stevan Harnad)
Date: Fri, 19 Aug 94 22:51:41 EDT
Subject: Brain Rhythms & Cognition: PSYCOLOQUY Call for Commentary
Message-ID: <9408200251.AA06967@clarity.Princeton.EDU>

  Pulvermueller et al: BRAIN RHYTHMS, CELL ASSEMBLIES AND COGNITION

The target article whose abstract appear below has just been just been
published in PSYCOLOQUY, a refereed electronic journal of Peer
Commentary sponsored by the American Psychological Association.
Formal commentaries are now invited. The full text can be easily and
instantly retrieved by a variety of simple means, described below.
Instructions for Commentators appear after the retrieval Instructions.

  TARGET ARTICLE AUTHOR'S RATIONALE FOR SOLICITING COMMENTARY:

  Fast periodic brain responses have been investigated in various
  mammals, humans included. Although most neuroscientists agree on the
  importance of these processes, it is not at all clear what role they
  play in cortical and subcortical processing. Are they simply a
  byproduct of perceptual processes, or do they play a role in what can
  be called higher or cognitive processing in the brain? We tried to
  answer this question by performing experiments in which spectral
  responses to meaningful words and physically similar but meaningless
  pseudowords were recorded from the human cortex. The result,
  differential 30-Hz responses to these stimuli, is interpreted in the
  framework of a Hebbian cell assembly theory. We hope that both the
  results and the brain-theoretic approach will stimulate a fruitful
  multidisciplinary discussion.

-----------------------------------------------------------------------
psycoloquy.94.5.48.brain-rhythms.1.pulvermueller  Friday 19 August 1994
ISSN 1055-0143   (30 paragraphs, 10 figs, 9 notes, 61 refs, 1203 lines)
PSYCOLOQUY is sponsored by the American Psychological Association (APA)
                Copyright 1994 Friedemann Pulvermueller et al.

        BRAIN RHYTHMS, CELL ASSEMBLIES AND COGNITION:
        EVIDENCE FROM THE PROCESSING OF WORDS AND PSEUDOWORDS

                Friedemann Pulvermueller (1)
                Hubert Preissl (1)
                Carsten Eulitz (2)
                Christo Pantev (2)
                Werner Lutzenberger (1)
                Thomas Elbert (2)
                Niels Birbaumer (1, 3)

    (1) Institut fuer Medizinische Psychologie und
        Verhaltensneurobiologie, Universitaet Tuebingen,
        Gartenstrasse 29, 72074 Tuebingen, Germany
        PUMUE at mailserv.zdv.uni-tuebingen.de

    (2) Institut fuer Experimentelle Audiologie, 
        Universitaet Muenster, Kardinal von Galen-Ring 10, 
        48149 Muenster, Germany

    (3) Universita degli Studi, Padova, Italy


    ABSTRACT: In modern brain theory, cortical cell assemblies are
    assumed to form the basis of higher brain functions such as form
    and word processing. When gestures or words are produced and
    perceived repeatedly by the infant, cell assemblies develop which
    represent these building blocks of cognitive processing. This leads
    to an obvious prediction: cell assembly activation ("ignition")
    should take place upon presentation of items relevant for cognition
    (e.g., words, such as "moon"), whereas no ignition should occur
    with meaningless items (e.g., pseudowords, such as "noom"). Cell
    assembly activity may be reflected by high-frequency brain
    responses, such as synchronous oscillations or rhythmic
    spatiotemporal activity patterns in which large numbers of neurons
    participate. In recent MEG and EEG experiments, differential
    gamma-band responses of the human brain were observed upon
    presentation of words and pseudowords. These findings are
    consistent with the view that fast coherent and rhythmic activation
    of large neuronal assemblies takes place with word but not
    pseudowords.

    KEYWORDS: brain theory, cell assembly, cognition, event related
    potentials (ERP), electroencephalograph (EEG), gamma band, Hebb,
    language, lexical processing, magnetoencephalography (MEG),
    psychophysiology, periodicity, power spectral analysis, synchrony


The article is retrievable from the following sites:

ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1994.volume.5
http://info.cern.ch/hypertext/DataSources/bySubject/Psychology/Psycoloquy.html
http://www.princeton.edu/~harnad/
gopher://gopher.cic.net/11/e-serials/alphabetic/p/psycoloquy
gopher://gopher.lib.virginia.edu/11/alpha/psyc
gopher://wachau.ai.univie.ac.at/11/archives/Psycoloquy

The filename is:
psycoloquy.94.5.48.brain-rhythms.1.pulvermueller

   To retrieve a file by ftp from a Unix/Internet site, type either:
ftp princeton.edu
   or
ftp 128.112.128.1
   When you are asked for your login, type:
anonymous
   Enter password as per instructions (your password is your own userid:
   yourlogin at yourhost.whatever.whatever - be sure to include the "@")
   and then change directories with:
cd pub/harnad/BBS
   To show the available files, type:
ls
   Next, retrieve the file you want with (for example):
get psycoloquy.94.5.48.brain-rhythms.1.pulvermueller
or
mget *pulv*
   When you have the file(s) you want, type:
quit
---------------------------------------------------------------- 
The file is also retrievable through archie, gopher, and World-Wide Web
using the URLs (Universal Resource Locators):

http://info.cern.ch/hypertext/DataSources/bySubject/Psychology/Psycoloquy.html
ftp://princeton.edu/pub/harnad/Psycoloquy
or
gopher://gopher.princeton.edu/1ftp%3aprinceton.edu%40/pub/harnad/BBS/
----------------------------------------------------------------
   Certain non-Unix/Internet sites have a facility you can use that is
   equivalent to the above. Sometimes the procedure for connecting to
   princeton.edu will be a two step process such as:
ftp
   followed at the prompt by:
open princeton.edu
  or
open 128.112.128.1
   In case of doubt or difficulty, consult your system manager.
------------------------------------------------------------------
Where the above procedures are not available (e.g. from Bitnet or other
networks), there are two fileservers --  ftpmail at decwrl.dec.com and
bitftp at pucc.bitnet -- that will do the transfer for you. Send either one the
one line message:
help
for instructions (which will be similar to the above, but will be in
the form of a series of lines in an email message that ftpmail or
bitftp will then execute for you).

--------------------------------------------------------------------------
        INSTRUCTIONS FOR PSYCOLOQUY AUTHORS AND COMMENTATORS

PSYCOLOQUY is a refereed electronic journal (ISSN 1055-0143) sponsored
on an experimental basis by the American Psychological Association
and currently estimated to reach a readership of 40,000. PSYCOLOQUY
publishes brief reports of new ideas and findings on which the author
wishes to solicit rapid peer feedback, international and
interdisciplinary ("Scholarly Skywriting"), in all areas of psychology
and its related fields (biobehavioral science, cognitive science,
neuroscience, social science, etc.). All contributions are refereed.

Target article length should normally not exceed 500 lines [c. 4500 words].
Commentaries and responses should not exceed 200 lines [c. 1800 words].

All target articles, commentaries and responses must have (1) a short
abstract (up to 100 words for target articles, shorter for commentaries
and responses), (2) an indexable title, (3) the authors' full name(s)
and institutional address(es).

In addition, for target articles only: (4) 6-8 indexable keywords,
(5) a separate statement of the authors' rationale for soliciting
commentary (e.g., why would commentary be useful and of interest to the
field? what kind of commentary do you expect to elicit?) and
(6) a list of potential commentators (with their email addresses).

All paragraphs should be numbered in articles, commentaries and
responses (see format of already published articles in the PSYCOLOQUY
archive; line length should be < 80 characters, no hyphenation).

It is strongly recommended that all figures be designed so as to be
screen-readable ascii. If this is not possible, the provisional
solution is the less desirable hybrid one of submitting them as
postscript files (or in some other universally available format) to be
printed out locally by readers to supplement the screen-readable text
of the article.

PSYCOLOQUY also publishes multiple reviews of books in any of the above
fields; these should normally be the same length as commentaries, but
longer reviews will be considered as well. Book authors should submit a
500-line self-contained Precis of their book, in the format of a target
article; if accepted, this will be published in PSYCOLOQUY together
with a formal Call for Reviews (of the book, not the Precis). The
author's publisher must agree in advance to furnish review copies to the
reviewers selected.

Authors of accepted manuscripts assign to PSYCOLOQUY the right to
publish and distribute their text electronically and to archive and
make it permanently retrievable electronically, but they retain the
copyright, and after it has appeared in PSYCOLOQUY authors may
republish their text in any way they wish -- electronic or print -- as
long as they clearly acknowledge PSYCOLOQUY as its original locus of
publication. However, except in very special cases, agreed upon in
advance, contributions that have already been published or are being
considered for publication elsewhere are not eligible to be considered
for publication in PSYCOLOQUY,

Please submit all material to psyc at pucc.bitnet or psyc at pucc.princeton.edu
Anonymous ftp archive is DIRECTORY pub/harnad/Psycoloquy HOST princeton.edu

From miku at sedal.su.oz.au  Tue Aug 23 03:02:45 1994
From: miku at sedal.su.oz.au (Michael Usher)
Date: Tue, 23 Aug 1994 17:02:45 +1000 (EST)
Subject: Sixth Australian Conference on Neural Networks
Message-ID: <199408230702.RAA26903@sedal.sedal.su.OZ.AU>

For those people putting the finishing touches to their submissions,
the Author's Style Guidelines are now available via the World Wide Web.
LaTeX style information can be found at URL:
	http://www.sedal.su.oz.au/acnn95

The conference programme and registration details will also be placed there,
when they become available.

Further queries about ACNN'95 should be directed to <acnn95 at ee.su.oz.au>.
Problems with the WWW server should be directed to myself.

Michael Usher
--
Michael Usher		Systems Administrator
miku at sedal.su.OZ.AU	Systems Engineering & Design Automation Lab (SEDAL)
Tel: +61 2 692 4135	Department of Electrical Engineering, Building J03
Fax: +61 2 660 1228	University of Sydney, NSW 2006, AUSTRALIA

From shultz at hebb.psych.mcgill.ca  Tue Aug 23 09:10:34 1994
From: shultz at hebb.psych.mcgill.ca (Tom Shultz)
Date: Tue, 23 Aug 94 09:10:34 EDT
Subject: No subject
Message-ID: <9408231310.AA19055@hebb.psych.mcgill.ca>

Subject: Paper available: A connectionist model of the learning of 
personal pronouns in English.
Date: 23 August '94

FTP-host: archive.cis.ohio-state.edu
FTP-file: pub/neuroprose/shultz.pronouns.ps.Z

-------------------------------------------------------------

The following paper has been placed in the Neuroprose archive at 
Ohio State University: 

A connectionist model of the learning of personal pronouns in 
English. (13 pages)

Thomas R. Shultz, David Buckingham, & Yuriko Oshima-Takane
Department of Psychology & McGill Cognitive Science Centre		
McGill University
Montreal, Quebec, Canada H3A 1B1		
shultz at psych.mcgill.ca

Abstract
Both experimental and observational psycholinguistic research have 
shown that children's acquisition of first and second person 
pronouns is affected by the opportunity to hear these pronouns used 
in speech not addressed to them. These effects were simulated with 
the cascade-correlation connectionist algorithm. The networks 
learned, in effect, to produce the pronouns "me" and "you" depending 
on identification of the speaker, addressee, and referent. Analysis of 
network performance and structure indicated that generalization to 
correct pronoun production was aided by listening to non-addressed 
speech and that persistent pronoun errors were created by listening 
to directly addressed speech. It was noted that explicit symbolic 
rule models would likely have difficulty simulating the pattern 
frequency effects common to the present simulations and to the 
natural language environment of the child.

The paper has been published in S. J. Hanson, T. Petsche, M. Kearns, & 
R. L. Rivest (Eds.) (1994). Computational learning theory and natural 
learning systems, Vol. 2: Intersection between theory and 
experiment (pp. 347-362). Cambridge, MA: MIT Press.  

Instructions for ftp retrieval of this paper are given below.  If you 
are unable to retrieve and print it and therefore wish to receive a 
hardcopy, please send e-mail to shultz at psych.mcgill.ca

Please do not reply directly to this message.

FTP INSTRUCTIONS:

unix> ftp archive.cis.ohio-state.edu (or 128.146.8.52)
    Name: anonymous
    Password: <your e-mail address>
    ftp> cd pub/neuroprose
    ftp> binary
    ftp> get shultz.pronouns.ps.Z 
    ftp> quit
unix> uncompress shultz.pronouns.ps.Z 

Thanks to Jordan Pollack for maintaining this archive. 

Tom Shultz

From tap at cs.toronto.edu  Tue Aug 23 16:18:48 1994
From: tap at cs.toronto.edu (Tony Plate)
Date: Tue, 23 Aug 1994 16:18:48 -0400
Subject: Thesis available for ftp
Message-ID: <94Aug23.161849edt.240@neuron.ai.toronto.edu>


Ftp-host: ftp.cs.utoronto.ca
Ftp-filename: /pub/tap/plate.thesis.2up.ps.Z
Ftp-filename: /pub/tap/plate.thesis.ps.Z

The following thesis is available for ftp.  There are two
versions: plate.thesis.ps.Z prints on 216 sheets of
paper, and plate.thesis.2up.ps.Z prints on 108 sheets.
The compressed files are around 750Kb each.

	Distributed Representations and
	Nested Compositional Structure

		    by

		Tony A. Plate*
	Department of Computer Science,
	   University of Toronto,
		   1994

A thesis submitted in conformity with the requirements for
the degree of Doctor of Philosophy in the Graduate Department
of Computer Science at the University of Toronto.


		 Abstract

Distributed representations are attractive for a number of
reasons.  They offer the possibility of representing
concepts in a continuous space, they degrade gracefully with
noise, and they can be processed in a parallel network of
simple processing elements.  However, the problem of
representing nested structure in distributed representations
has been for some time a prominent concern of both
proponents and critics of connectionism
\cite{fodor-pylyshyn-88,smolensky-90,hinton-90}.  The lack
of connectionist representations for complex structure has
held back progress in tackling higher-level cognitive tasks
such as language understanding and reasoning.

In this thesis I review connectionist representations and
propose a method for the distributed representation of
nested structure, which I call ``Holographic Reduced
Representations'' (HRRs).  HRRs provide an implementation of
Hinton's~\shortcite{hinton-90} ``reduced descriptions''.
HRRs use circular convolution to associate atomic items,
which are represented by vectors.  Arbitrary variable
bindings, short sequences of various lengths, and
predicates can be represented in a fixed-width vector.
These representations are items in their own right, and can
be used in constructing compositional structures.  The noisy
reconstructions extracted from convolution memories can be
cleaned up by using a separate associative memory that has
good reconstructive properties.

Circular convolution, which is the basic associative
operator for HRRs, can be built into a recurrent neural
network.  The network can store and produce sequences.  I
show that neural network learning techniques can be used
with circular convolution in order to learn representations
for items and sequences.

One of the attractions of connectionist representations of
compositional structures is the possibility of computing
without decomposing structures.  I show that it is possible
to use dot-product comparisons of HRRs for nested structures
to estimate the analogical similarity of the structures.
This demonstrates how the surface form of connectionist
representations can reflect underlying structural similarity
and alignment.


* New Address:

Tony Plate,
Department of Chemical Engineering
University of British Columbia
2216 Main Mall, Vancouver,
BC, Canada V6T 1Z4
tap at chml.ubc.ca

From d.rosen at ieee.org  Wed Aug 24 10:47:12 1994
From: d.rosen at ieee.org (d.rosen@ieee.org)
Date: Wed, 24 Aug 94 07:47:12 -0700
Subject: Software Engineer/Developer (N.Y.) [Neural Networks] Wanted
Message-ID: <9408241445.AA14964@solstice>

Full-time position for Software Engineer/Developer with recent Degree.
Possible alternative: Part-time Experienced Software Engineer/Developer.

Starting date: Immediate/September.

A federally-funded research team at New York Medical College is
applying neural networks and advanced probabilistic/statistical
methods to improve the accuracy with which the stage (of advancement)
of cancer cases can be evaluated -- an important factor in determining
treatment.

We seek a skilled developer to take primary responsibility for the
design and implementation of our neural network software, which will
be geared towards flexible experimental use in our fast-paced research
program, as well as a simple GUI prototype of clinical production
software.  The successful candidate will work with us (medical, neural
network, and statistical researchers) to plan the best path from our
current C code to a more carefully-designed, extensible, OO approach,
perhaps with partial rapid implementation in an interpreted language,
and evaluate the possible role of other available tools and libraries.
It is expected that eventually, much of the resulting software will be
freely distributed for use in many fields.

Candidates should have demonstrably outstanding skills in designing
object-oriented software, and in C++ (or both C and some OO language)
development under the Unix[/X11] environment.  Prefer knowledge of as
many of the following as possible: neural nets, statistics, numerical
/ scientific computation, portable GUI, prototyping language (Python,
Smalltalk, Perl, S-plus, ...), MS Windows, and Unix system
administration.

New York Medical College (NYMC) is located in the community of
Valhalla, NY, just half an hour north of New York City.  The position
would be on-site, though doing some portion of the work remotely could
perhaps be arranged.

NYMC is the third-largest private medical university in the United
States. It is an Equal Opportunity and Affirmative Action Institution.
Currently we do not believe we could justify hiring an individual who
is not already authorized to work in the U.S.

If you are interested and qualified, please e-mail your resume to me
as soon as possible (plain text preferred).

--
David Rosen, PhD   <d.rosen at ieee.org>

From iiscorp at netcom.com  Wed Aug 24 18:48:07 1994
From: iiscorp at netcom.com (IIS Corp)
Date: Wed, 24 Aug 94 15:48:07 PDT
Subject: Soft Computing Days in San Francisco
Message-ID: <CMM.0.90.4.777768487.iiscorp@netcom>



              Soft Computing Days in San Francisco


           Zadeh, Widrow, Koza, Ruspini, Stork, Whitley, 
                  Bezdek, Bonissone, and Berenji

                              On

              
                         Soft Computing:  
	 Fuzzy Logic, Neural Networks, and Genetic Algorithms

     
                      Three short courses


                       San Francisco, CA
                       October 24-28, 1994


Traditional (hard) computing methods do not provide sufficient
capabilities to develop and implement intelligent systems.  Soft
computing methods have proved to be important practical tools to build
and construct these systems.  

The following three courses, offered by Intelligent Inference Systems 
Corp., will focus on all major soft computing technologies: fuzzy logic,
neural networks, genetic algorithms, and genetic programming. These
courses may be taken either individually or in combination.

Course 1: Artificial Neural Networks (Oct. 24)
          Bernard Widrow and David Stork

Course 2: Genetic Algorithms and Genetic Programming (Oct. 25)
	  John Koza and Darrell Whitley

Course 3: Fuzzy Logic Inference (Oct. 26-28)
	  Lotfi Zadeh, Jim Bezdek, Enrique Ruspini, 
	  Piero Bonissone, and Hamid Berenji

For further details on course topics and registration information, send
an email to iiscorp at netcom.com or contact Intelligent Inference
Systems Corp., Phone (415) 988-9934, Fax: (415) 988-9935.  A detailed
brochure will be sent to you as soon as possible.


From ken at phy.ucsf.edu  Thu Aug 25 08:16:49 1994
From: ken at phy.ucsf.edu (ken@phy.ucsf.edu)
Date: Thu, 25 Aug 1994 05:16:49 -0700
Subject: postdoctoral position available in computational neuroscience
Message-ID: <9408251216.AA20136@coltrane.ucsf.EDU>

Christof Schreiner works on the physiology of auditory cortex.  His
lab has defined several of the auditory parameters that seem to be
mapped in auditory cortex.  He has funding for a postdoctoral position
for theoretical work aimed at understanding the mapping to auditory
cortex and its possible consequences for representation of complex
sounds.  His description follows:

	The objective is to develop network models of the mammalian auditory
	cortex on the basis of a broad data base of physiological observations
	from our laboratory. Based on the spectral-temporal filter properties
	of neurons and their spatial distribution in the cortex, consequences
	for the cortical representation of complex signals (such as animal
	vocalizations and speech) shall be evaluated. Previous experience in
	signal processing and self-organizing or other neural network classes
	is required.

Christof and I are both members of the Keck Center for Integrative
Neuroscience at UCSF, a group of 10 faculty working on systems
neuroscience including Michael Stryker, Michael Merzenich, Steve
Lisberger, Allison Doupe, Alan Basbaum, Roger Nicoll, Howard Fields,
and Henry Ralston.  The postdoc will work in the Keck Center, and
both I and Christof will be available to work closely with them.

DO NOT SEND YOUR APPLICATIONS TO ME!!

Send applications to:

Christof Schreiner
Dept. of Otolaryngology
Box 0732
UCSF
513 Parnassus
SF, CA 94143-0732

email: chris at phy.ucsf.edu

Please send a cv, and names, addresses and phone numbers of three
individuals who can provide references for you.  Copies of your
publications would also be helpful.

Ken Miller

From bishopc at helios.aston.ac.uk  Thu Aug 25 07:18:00 1994
From: bishopc at helios.aston.ac.uk (bishopc)
Date: Thu, 25 Aug 1994 11:18:00 +0000
Subject: NIPS Workshop - Inverse Problems
Message-ID: <4310.9408251018@sun.aston.ac.uk>


-------------------------------------------------------------------

                      NIPS*94 Workshop:


                     DOING IT BACKWARDS: 
                     -------------------
 
      NEURAL NETWORKS AND THE SOLUTION OF INVERSE PROBLEMS
      ----------------------------------------------------
 
                  Organizer: Chris M Bishop



                    CALL FOR PRESENTATIONS
                    ---------------------- 


Introduction:
------------

Many of the tasks for which neural networks are commonly used 
correspond to the solution of an `inverse' problem. Such tasks are 
characterized by the existence of a well-defined, deterministic 
`forward' problem which might, for instance, correspond to causality 
in a physical system. By contrast the inverse problem may be 
ill-posed, and may exhibit multiple solutions. 


Aims:
----

A wide range of different approaches have been developed to tackle
inverse problems, and one of the main goals of the workshop is to 
contrast the way in which they address the underlying technical issues, 
and to identify key areas for future research.


Format:
------

This will be a one-day workshop, and will involve short presentations,
with ample time allowed for discussions. 


Contributions
-------------

If you wish to make a contribution to this workshop, please e-mail a 
brief outline of your proposed presentation to c.m.bishop at aston.ac.uk 
by 5 September.


Chris Bishop


-------------------------------------------------------------------- 

  Professor Chris M Bishop           Tel. +44 (0)21 359 3611 x4270
  Neural Computing Research Group    Fax. +44 (0)21 333 6215
  Dept. of Computer Science          c.m.bishop at aston.ac.uk
  Aston University               
  Birmingham B4 7ET, UK

--------------------------------------------------------------------


From oby at TechFak.Uni-Bielefeld.DE  Thu Aug 25 13:13:10 1994
From: oby at TechFak.Uni-Bielefeld.DE (oby@TechFak.Uni-Bielefeld.DE)
Date: Thu, 25 Aug 94 19:13:10 +0200
Subject: No subject
Message-ID: <9408251713.AA02980@gaukler.TechFak.Uni-Bielefeld.DE>


POSTDOCTORAL POSITION

Postdoctoral position available beginning spring / summer 1995 in a project
combining anatomy, physiology and theoretical modelling to understand the
functional architecture of primate visual cortex. Funded by the Human
Frontiers program, the position is initially for two years and the
postdoctoral fellow will be expected to interact with members of the group
in Germany, England, and USA.


The candidate should have some experience with neural modeling and possess
appropriate math and and computer competency; the candidate should also
have relevant experimental skills (eg anatomy or physiological unit
recording) in mammalian visual system.


Applicants should send their CV, list of publications, letter describing
their interest in the position, and name,address and phone number of two
referees to either


Prof. Jennifer Lund, 

Institute of Ophthalmology, 

11-43 Bath St. , London EC1V 9EL,UK 

Phone (0)71-608-6864), E-mail smgxjsl at ucl.ac.uk; 


or 


Dr. Klaus Obermayer, 

Universitaet Bielefeld, Technische Fakultat,
Universitaetsstrasse 25, 33615 Bielefeld, Germany 

Phone (0)521-106-6058; Email oby at techfak.uni-bielefeld.de.


From ccg at melissa.lanl.gov  Thu Aug 25 18:23:12 1994
From: ccg at melissa.lanl.gov (Camilo Gomez)
Date: Thu, 25 Aug 94 16:23:12 MDT
Subject: positions in financial research
Message-ID: <9408252223.AA00343@melissa.lanl.gov.demosdiscs>

(PLEASE POST)

-----------------------------------------------------------------------------------------

               POSITIONS IN FINANCIAL RESEARCH
                             AT
               LOS ALAMOS NATIONAL LABORATORY
                CENTER FOR NON-LINEAR STUDIES
             (Postdoctoral and Graduate Student)

Positions  involving research and development  of  financial
models  are  available  at Los Alamos  National  Laboratory.
Depending  on  funding we will have a  number  of  positions
including:

     a)postdoctoral
     b)graduate student

The  successful  candidates will  be  expected  to  work  on
projects involving research and development in the  area  of
financial  derivatives. These will involve both fixed-income
and  equity  derivatives. Projects will focus  on  valuation
problems  for  a  number  of  these  financial  derivatives.
Exceptionally well qualified candidates with an interest  in
computational investigations of above mentioned  topics  and
expertise in one or more of the following or related  areas,
are encouraged to apply:

     a)finance
     b)financial derivatives
     c)statistical analysis
     d)time series analysis
     e)neural net/pattern recognition
     f)emergent behavior systems
     g)parallel computing
     h)programming skills (C and C++ languages)

Candidates may contact:


     M.F. Gomez
     CNLS, MS-B258
     Los Alamos National Laboratory
     Los Alamos, NM 87544
     frankie at cnls.lanl.gov

for application material and questions.

Please indicate in your initial inquiry that is for a
position in financial research and whether you are
interested in a student or post-doctoral position.

Los Alamos National Laboratory is an equal opportunity
employer.

-----------------------------------------------------------------------------------------

From shannon at cis.ohio-state.edu  Thu Aug 25 15:02:52 1994
From: shannon at cis.ohio-state.edu (shannon roy campbell)
Date: Thu, 25 Aug 1994 15:02:52 -0400 (EDT)
Subject: paper available on synchrony and desynchrony in oscillator an oscillator network
Message-ID: <199408251902.PAA27060@axon.cis.ohio-state.edu>


FTP-host: archive.cis.ohio-state.edu
FTP-filename : /pub/neuroprose/campbell.wc_oscillators.ps.Z

The file campbell.wc_oscillators.ps.Z (34 pages) is now 
available for copying from the Neuroprose repository. 
Contact shannon at cis.ohio-state.edu for retrieval problems.

          
           Synchronization and Desynchronization in a 
        Network of Locally Coupled Wilson-Cowan Oscillators

            by Shannon Campbell and DeLiang Wang~
               Dept. of Physics 
              ~Dept. of Computer and Information Science
               Ohio State University
               Columbus, Ohio 43210-1277

Abstract - 
   A network of Wilson-Cowan oscillators is constructed, and its
emergent properties of synchronization and desynchronization are 
investigated by both computer simulation and formal analysis. The 
network is a two-dimensional matrix where each oscillator is
coupled only to its neighbors. We show analytically that a chain of
locally coupled oscillators (the piece-wise linear approximation to 
the Wilson-Cowan oscillator) synchronizes, and present a technique 
to rapidly entrain finite numbers of oscillators. The coupling
strengths change on a fast time scale based on a Hebbian rule. A
global separator is introduced which receives input from and sends 
feedback to each oscillator in the matrix. The global separator is 
used to desynchronize different oscillator groups. Unlike many 
other models, the properties of this network emerge from local 
connections, that preserve spatial relationships among components 
and are critical for encoding Gestalt principles of feature 
grouping. The ability to synchronize and desynchronize oscillator
groups within this network offers a promising approach for pattern 
segmentation and figure/ground segregation based on oscillatory
correlation. 

FTP INSTRUCTIONS:

unix> ftp archive.cis.ohio-state.edu (or 128.146.8.52)
    Name: anonymous
    Password: <your e-mail address>
    ftp> cd pub/neuroprose
    ftp> binary
    ftp> get campbell.wc_oscillators.ps.Z 
    ftp> quit
unix> uncompress campbell.wc_oscillators.ps.Z 

From shannon at cis.ohio-state.edu  Fri Aug 26 10:25:48 1994
From: shannon at cis.ohio-state.edu (shannon roy campbell)
Date: Fri, 26 Aug 1994 10:25:48 -0400 (EDT)
Subject: correct reference for paper on synchrony and desynchrony...
Message-ID: <199408261425.KAA27616@axon.cis.ohio-state.edu>


        Sorry that I did not specify the proper reference to a previously
        announced paper in neuroprose ("Synchrony and desynchrony in
        a network of Wilson-Cowan oscillators"). The paper is a techical
        report from The Department of Computer and Information Science,
        The Ohio State University, numbered "OSU-CISRC-8/94-TR43".

        Thanks for your attention.
        -shannon campbell

From kreider at bechtel.Colorado.EDU  Sat Aug 27 16:21:38 1994
From: kreider at bechtel.Colorado.EDU (Jan Kreider)
Date: Sat, 27 Aug 1994 14:21:38 -0600
Subject: System identification competition announcement
Message-ID: <199408272021.OAA10216@bechtel.Colorado.EDU>


Announcement of 

***System Identification Competition***

Benchmark tests for estimation methods of thermal characteristics of 
buildings and building components.


*Objective

The objective of the benchmark is to set-up a comparison between 
alternative techniques and to clarify particular problems of system 
identification applied to the thermal performance of buildings. 


*Organisation

J. Bloem, U. Norlen, EC - Joint Research Centre, Ispra, Italy
H. Madsen, H. Melgaard, IMM TU of Denmark, Lyngby, Denmark
J. Kreider, JCEM, University of Colorado, U.S.


*Active period of the competition : July 1, 1994 - December 31, 1994


*Introduction

A wide range of system identification techniques is now being applied 
to the analysis problems involved with estimation of thermal properties 
of buildings and building components. Similar problems arise in most 
observational disciplines, including physics, biology, and economics. 
New commercially available software tools and special purpose computer 
programs promise to provide results that were unobtainable just a 
decade ago. Unfortunately, the realisation and evaluation of this 
promise has been hampered by the difficulty of making rigorous 
comparisons between competing techniques, particularly ones that come 
from different disciplines.

This competition has been organised to help clarify the conflicting 
claims among many researchers who use and analyse building energy data 
and to foster contact among these persons and their institutions. The 
intent is not necessarily only to declare winners but rather to set up 
a format in which rigorous evaluations of techniques can be made. 
Because there are natural measures of performance, a rank-ordering will 
be given. In all cases, however, the goal is to collect and analyse 
quantitative results in order to understand similarities and differences 
among the approaches.

At the close of the competition the performance of the techniques 
submitted will be compared.  Those with the best results will be asked 
to write a scientific paper and will be invited for a presentation of 
the paper. There will be no monetary prizes. A symposium at the JRC 
Ispra, Northern Italy, has been scheduled for the Autumn 1995 to 
explore the results of the competition in formal papers. The competition,
 the overall results and papers on selected methods will published by 
the organisers in a book.

Research on energy savings in buildings can be divided in three major 
areas:

1) building components, 2) test cells and unoccupied buildings in 
real climate and 3) occupied buildings. Three competitions are planned 
along this line of which the present competition concerned with building 
components will be the first one.

The present competition is concerned with wall components and no solar 
radiation involved. Five different cases are provided for estimation and
prediction. Four cases have been designed with wall components in order
to test parameter estimation methods. Prediction tests are also 
included. Some of the dependent variable values will be withheld from 
the data set in these cases. Contestants are free to submit results 
from any number of cases.

When the outcome of this first competition is positive a second 
competition is planned which concerns test cells and unoccupied 
buildings under real climate conditions (1995). A third competition 
concerns occupied buildings (1996).

If there is sufficient interest, a network server may be set up to 
operate as an on-line archive of interesting data sets, programs, and 
comparisons among algorithms in the future. 


***ACCESSING THE DATA

The competition does not require advanced registration; there are two 
ways to enter: 

1. by normal mail. Simply request the data by sending a letter. The 
data are available on diskettes (3.5-in size) in ASCII, IBM-PC format. 
(there is no charge for the data diskette). To receive the data, 
send the letter together with a self-addressed rigid envelope to :

Joint Research Centre
Institute of System Engineering and Informatics
J.J. BLOEM,     Building 45
I - 21020 Ispra (VA), ITALY

2. by E-mail. In that case just send an request for participation by 
E-mail to J. Kreider at the University of Colorado, Boulder, 
CO 80309-0428, U.S. at the following E-mail address: 

jkreider at vaxf.colorado.edu

Information how to obtain the necessary instructions and the required 
data series, using FTP, are forwarded to you by E-mail.
Instructions on submitting a return disk with the analysis of the cases 
will be included in a README file. The disk will also include an entry 
form that each participant will need to complete and submit along with 
the results.

***FOR MORE INFORMATION


Further questions about the competition should be directed to one of 
the following organisers:


Joint Research Centre
Institute for Systems Engineering 
and Informatics
J.J. BLOEM      Building 45
I - 21020 ISPRA (VA), Italy
tel:    +39 332 789842/789145
fax:    +39 332 789992
E-mail:  hans.bloem at cen.jrc.it


Joint Center for Energy Management
J. KREIDER
Campus Box 428
University of Colorado
Boulder, CO  80309-0428, U.S.
tel:    +303 492 3915
fax:    +303 492 7317
E-mail:  jkreider at vaxf.colorado.edu


End of Announcement SysId Competition.


From lautrup at connect.nbi.dk  Mon Aug 29 15:54:23 1994
From: lautrup at connect.nbi.dk (Benny Lautrup)
Date: Mon, 29 Aug 94 15:54:23 METDST
Subject: no subject (file transmission)
Message-ID: <mailman.700.1149591315.29955.connectionists@cs.cmu.edu>

Subject: Paper available: Extremely Ill-posed Learning
Date: August 29, 1994

FTP-host: connect.nbi.dk
FTP-file: neuroprose/hansen.ill-posed.ps.Z

----------------------------------------------

The following paper is now available:

Extremely Ill-posed Learning [14 pages]

L.K. Hansen, B. Lautrup, I. Law, N. Moerch, and J. Thomsen
CONNECT, The Niels Bohr Institute, University of Copenhagen, Denmark
 
Abstract:
Extremely ill-posed learning problems  are  common  in  image  and
spectral analysis. They are characterized  by  a  vast  number  of
highly correlated inputs, eg pixel  or  or  pin  values,  and  a
modest number of patterns, eg images or spectra. We show that it
is possible to  train  neural  networks  to  learn  such  patterns
without using an excessive number of weights, and we devise a test
to decide if new patterns should be included in the  training  set
or whether they fall within the  subspace  already  explored.  The
method is applied to the analysis of PET-images.

Please do not reply directly to this message.

-----------------------------------------------

FTP-instructions:

unix> ftp connect.nbi.dk (or 130.225.212.30)
ftp> Name: anonymous
ftp> Password: your e-mail address
ftp> cd neuroprose
ftp> binary
ftp> get hansen.ill-posed.ps.Z
ftp> quit
unix> uncompress hansen.ill-posed.ps.Z
        
-----------------------------------------------

Benny Lautrup,
Computational Neural Network Center (CONNECT)
Niels Bohr Institute
Blegdamsvej 17
2100 Copenhagen
Denmark

Telephone: +45-3532-5200
Direct:    +45-3532-5358
Fax:       +45-3142-1016
e-mail:    lautrup at connect.nbi.dk

From marshall at cs.unc.edu  Mon Aug 29 12:03:50 1994
From: marshall at cs.unc.edu (Jonathan A. Marshall)
Date: Mon, 29 Aug 1994 12:03:50 -0400
Subject: POSTDOC JOB (Neural net models of learning)
Message-ID: <199408291603.MAA27461@marshall.cs.unc.edu>

**********************************************************************

POSTDOCTORAL POSITION 

Available in September 1994.  Focus of research is neural network
models of learning, with a special emphasis on spatial tasks.

Qualified Ph.D. applicants should send a cover letter indicating
research experience, vitae, and names of three references to

Dr. Nestor Schmajuk
Department of Psychology
Duke University
Durham, NC 27706

e-mail: nestor at acpub.duke.edu

**********************************************************************

From ccg at melissa.lanl.gov  Tue Aug 30 11:01:01 1994
From: ccg at melissa.lanl.gov (Camilo Gomez)
Date: Tue, 30 Aug 94 09:01:01 MDT
Subject: positions in financial research (corrected e-mail)
Message-ID: <9408301501.AA02500@melissa.lanl.gov.demosdiscs>

(PLEASE POST)
-----------------------------------------------------------------------------------------

               POSITIONS IN FINANCIAL RESEARCH
                             AT
               LOS ALAMOS NATIONAL LABORATORY
                CENTER FOR NON-LINEAR STUDIES
             (Postdoctoral and Graduate Student)

Positions  involving research and development  of  financial
models  are  available  at Los Alamos  National  Laboratory.
Depending  on  funding we will have a  number  of  positions
including:

     a)postdoctoral
     b)graduate student

The  successful  candidates will  be  expected  to  work  on
projects involving research and development in the  area  of
financial  derivatives. These will involve both fixed-income
and  equity  derivatives. Projects will focus  on  valuation
problems  for  a  number  of  these  financial  derivatives.
Exceptionally well qualified candidates with an interest  in
computational investigations of above mentioned  topics  and
expertise in one or more of the following or related  areas,
are encouraged to apply:

     a)finance
     b)financial derivatives
     c)statistical analysis
     d)time series analysis
     e)neural net/pattern recognition
     f)emergent behavior systems
     g)parallel computing
     h)programming skills (C and C++ languages)

Candidates may contact:


     M.F. Gomez
     CNLS, MS-B258
     Los Alamos National Laboratory
     Los Alamos, NM 87544
     frankie at goshawk.lanl.gov

for application material and questions.

Please indicate in your initial inquiry that is for a
position in financial research and whether you are
interested in a student or post-doctoral position.

Los Alamos National Laboratory is an equal opportunity
employer.

-----------------------------------------------------------------------------------------

From prechelt at ira.uka.de  Tue Aug 30 15:32:35 1994
From: prechelt at ira.uka.de (Lutz Prechelt)
Date: Tue, 30 Aug 1994 21:32:35 +0200
Subject: Report on NN learning algorithm evaluation practices
Message-ID: <"irafs2.ira.198:30.07.94.19.32.14"@ira.uka.de>


A technical report titled

  A Study of Experimental Evaluations
  of Neural Network Learning Algorithms:
  Current Research Practice

is now available for anonymous ftp (binary mode!) from

  archive.cis.ohio-state.edu  /pub/neuroprose/prechelt.eval.ps.Z

The file is 70 Kb, the paper has 9 pages.

This is the abstract:

113 journal articles about neural network learning algorithms published in
1993 and 1994 are examined for the amount of experimental evaluation they
contain. Every third of them does employ not even a single realistic or real
learning problem. Only 6\% of all articles present results for more
than one problem using real world data.
Furthermore, one third of all articles does not present any
quantitative comparison with a previously known algorithm.
These results indicate that the quality of research in the
area of neural network learning algorithms needs improvement.
The publication standards should be raised and easily accessible
collections of example problems be built.

Here is a bibtex entry for the report:

@techreport{Prechelt94d,
   author    = {Lutz Prechelt},
   title     = {A Study of Experimental Evaluations of Neural Network
                Learning Algorithms: Current Research Practice},
   institution = {Fakult\"at f\"ur Informatik, Universit\"at Karlsruhe},
   year      = {1994},
   number    = {19/94},
   address   = {D-76128 Karlsruhe, Germany},
   month     = aug,
   note      = {anonymous ftp: /pub/papers/techreports/1994/1994-19.ps.Z
                on ftp.ira.uka.de},
}

Comments are welcome.

  Lutz
  
Lutz Prechelt   (email: prechelt at ira.uka.de)            | Whenever you 
Institut fuer Programmstrukturen und Datenorganisation  | complicate things,
Universitaet Karlsruhe;  76128 Karlsruhe;  Germany      | they get
(Voice: ++49/721/608-4068, FAX: ++49/721/694092)        | less simple.

From tgd at chert.CS.ORST.EDU  Tue Aug 30 19:07:33 1994
From: tgd at chert.CS.ORST.EDU (Tom Dietterich)
Date: Tue, 30 Aug 94 16:07:33 PDT
Subject: paper: Error Correcting Output Codes
Message-ID: <9408302307.AA13478@edison.CS.ORST.EDU>

The following paper is available at URL:
"ftp://ftp.cs.orst.edu/pub/tgd/papers/tr-ecoc.ps.gz" 

               Solving Multiclass Learning Problems via
                    Error-Correcting Output Codes

                Thomas G. Dietterich  tgd at cs.orst.edu
                   Department of Computer Science,
                          303 Dearborn Hall
                       Oregon State University
                       Corvallis, OR 97331 USA

                            Ghulum Bakiri
                    Department of Computer Science
                        University of Bahrain
                          Isa Town, Bahrain


    Multiclass learning problems involve finding a definition for an
    unknown function f(x) whose range is a discrete set containing k>2
    values (i.e., k ``classes'').  The definition is acquired by studying
    large collections of training examples of the form <x_i, f(x_i)>.
    Existing approaches to multiclass learning problems include (a) direct
    application of multiclass algorithms such as the decision-tree
    algorithms C4.5 and CART, (b) application of binary concept learning
    algorithms to learn individual binary functions for each of the $k$
    classes, and (c) application of binary concept learning algorithms
    with distributed output representations such as those employed by
    Sejnowski and Rosenberg in the NETtalk system.  This paper compares
    these three approaches to a new technique in which error-correcting
    codes are employed as a distributed output representation.  We show
    that these output representations improve the generalization
    performance of both C4.5 and backpropagation on a wide range of
    multiclass learning tasks.  We also demonstrate that this approach is
    robust with respect to changes in the size of the training sample, the
    assignment of distributed representations to particular classes, and
    the application of overfitting avoidance techniques such as
    decision-tree pruning.  Finally, we show that--like the other
    methods--the error-correcting code technique can provide reliable
    class probability estimates.  Taken together, these results
    demonstrate that error-correcting output codes provide a
    general-purpose method for improving the performance of inductive
    learning programs on multiclass problems.


Thomas G. Dietterich              Voice: 503-737-5559
Department of Computer Science    FAX:   503-737-3014
Dearborn Hall, 303                URL:   http://www.cs.orst.edu/~tgd/index.html
Oregon State University
Corvallis, OR 97331-3102

From georgiou at wiley.csusb.edu  Wed Aug 31 14:00:01 1994
From: georgiou at wiley.csusb.edu (georgiou@wiley.csusb.edu)
Date: Thu, 1 Sep 1994 02:00:01 +0800
Subject: JCIS: Last Call for Papers
Message-ID: <9409010900.AA04051@orion.csci.csusb.edu>

	       JOINT CONFERENCE ON INFORMATION SCIENCES

			 Last Call for Papers

(For submission of Neural Networks papers please see end of message for
address.  Deadline: September 10, 1994)


ORGANIZERS 

Honorary Chairs                   
Lotfi A. Zadeh & Azriel Rosenfeld 

Managing Chair of the Joint Conferences 
Paul P. Wang 
Dept. of Electrical Engineering 
Duke University 
Durham, NC 27708-0291 
Tel: (919) 660 5271, 660-5259 
Fax: (919) 660-5293, 684-4860 
e-mail: ppw at ee.duke.edu 

Advisorv Board 
Nick Baker 
Earl Dowell 
Erol Gelenbe 
Stephen Grossberg 
Kaoru Hirota 
Abe Kandel 
George Klir 
Teuvo Kohonen 
Tosiyasu L. Kunii 
Jin-Cherng Lin 
E. Mamdani 
Xuan-Zhong Ni 
C.V Ramamoorthy 
John E.R Staddon 
Masaki Togai 
Victor Van Beuren 
Max Woodbury 
Stephen S. Yau 
Lotfi Zadeh 
H. Zimmerman 

Keynote Speakers 
Lotfi A. Zadeh & Stephen Grossberg 

Plenarv Speakers 
Suguru Arimoto 
Dennis Bahler 
James Bowen 
Abe Kandel 
George Klir 
Phillipe Smets 
John R. Rice 
l.B. Turksen 
Benjamin Wah 
Stephen S. Yau

General Information 

The JOINT CONFERENCE ON INFORMATION SCIENCES consists of two international
conferences and one workshop. All interested attendees including
researchers, organizers, speakers, exhibitors, students and other
participants should register either in Plan A: 3rd International Conference
on Fuzzy Theory and Technology or Plan B: First International Conference on
Computer Theory & lnformatics and Workshop on Mobile Computing Svstems. Any
participants can attend all the keynote speeches, plenary sessions, all
parallel sessions and exhibits. The only difference is that all authors
registered in Plan A will participate in Lotfi A. Zadeh BestPaper
Competition. Plan B will have no best paper competition; at least for this
year. In addition, each plan will publish its own proceeding.

Tutorials 

Session A: Fuzzy Theory & Technology
(Sunday, November 13, 1994, 8:30 am - 12:30 pm)

1. George  Klir .................................Fuzzy Set and  Logic 
2. l. B. Turksen ................................Fuzzy Expert Systems 
3. Jack Aldridge .......................................Fuzzy Control 
4. Marcus Thint .......................Fuzzy Logic and NN Integration

Session B: Computers 
(Sunday, November 13, 1994, 1:30pm- 6:30pm)

1. Richard Palmer ....................................Neural Network 
2. Frank Y. Shih ................................Pattern Recognition 
3. Patrick Wang ......Intelligent Pattern Recognition & Applications 
4., 5. Ken W. White...............Success with Machine Vision I & II

Time Schedule & Venue

Tutorials ....................November 13, 1994 o 8:30 am - 6:30 pm 
Conferences .............November 14 - 16, 1994 o 8:30 am - 5:30 pm 
Venue........Pinehurst Resort & Country Club, Pinehurst, NC, U.S.A.

PARTICIPATION PLAN A:
3rd Annual Conference on Fuzzy Theory and Technology

PROGRAM COMMITTEE

Jack Aldridge 
Suguru Arimoto 
W Bandler 
P Bonnisone 
Bruno Bosacchi 
B Bouchon-Meunier 
J Buckley 
Dev Garg 
Rhett George 
George Georgiou 
I R Goodman 
Siegfried Gottwald 
Silvia Guiasu 
M M. Gupta 
Ralph Horvath 
D L Hung 
Timothy Jacobs 
Y.K Jani 
Joaquim Jorge 
Paul Kainen 
S.C. Kak 
Abe Kandel 
P Klement 
L J Kohout 
Vladik Kreinovlch 
N Kuroki 
Reza Langari
Harry Hua Li 
Don Meyers 
C K Mitchell 
John Mordeson 
Akira Nakamura 
Kyung Whan Oh 
Maria Orlowska 
Robert M. Pap 
Arthur Rarner 
Elie Sanchez 
B Scwhott 
Shouyoe Shao 
Sujeet Shenoi 
Frank Shih 
H Allison Smith 
L M Sztandera 
Alade ToKuta 
R Tong 
I Turksen 
Guo Jun Wang 
Tom Whalen 
Edward K Wong 
T Yamakawa 

The conference will consist of both plenary sessions and contributory
sessions, focusing on topics of critical interest and the direction of
future research.  Example topics include, but are not limited to the
following:

TOPICS: 3rd ANNUAL CONFERENCE ON FUZZY THEORY AND TECHNOLoGY

o Fuzzy Mathematics
o Basic Principles and Foundations of Fuzzy Logic
o Qualitative and Approximate-Reasoning Modeling
o Hardware Implementations of Fuzzy Logic Algorithms
o Design . Analysis, and Synthesis of Fuzzy Logic Controllers
o Learning and Acquisition of Approximate Models
o Fuzzy Expert Systems          
o Neural Network Architectures
o Artificially Intelligent Neural Networks
o Artificial Life
o Associative Memory
o Computational Intelligence
o Cognitive Science
o Fuzzy Neural Systems
o Relations between Fuzzy Logic and Neural Networks
o Theory of Evolutionary Computation 
o Efficiency/robustness comparisons with other direct search algorithms
o Parallelcomputer:applications
o Integration of Fuzzy Logic and Evolutionary Computing
o Comparisons between different variants of evolutionary algorithms 
o Evolutionary Computation for neural networks
o Fuzzy logic in Evolutionary algorithms
o Neurocognition
o Neurodynamics
o Optirnization
o Pattern Recognition
o Learning and Memory
o Machine Learning Applications
o Implementations (electronic, Optical, biochips)
o Intelligent Control

APPLICATIONS OF THE TOPICS:

o Hybrid Systems
o Image Processing
o Image Understanding
o Pattern Recognition
o Robotics and Automation
o Intelligent Vehicle and Highway Systems 
o Virtual Reality
o Tactile Sensors 
o Machine Vision 
o Motion Analysis
o Nuro biology
o Sensation and Perception
o Sensorimotor Systems
o Speech, Hearing and Language 
o Signal Processing
o Time Series Analysis
o Prediction
o System ldentification  
o System Control
o Intelligent Information Systems
o Case-Based Reasoning 
o Decision Analysis 
o Databases and Information Retrieval
o Dynamic Systems Modeling & Diagnosis
o Electric & Nuclear Power Systems

PARTICIPATION PLAN B:
1st Annual Computer Theory and Informatics Conference
& Workshop on Mobile Computing Systems

PROGRAM COMMITTEE:
FIRST ANNUAL, COMPUTER & INFORMATICS

Rafi Ahmed 
R Alonso 
Suguru Arimoto 
B. Badrinath 
C.R. Baker 
Martin Boogaard 
O Bukhres 
P Chrysantis 
Eliseo Clementini 
E.M. Ehlers 
Ahmed Elmagarmid 
K Ferentinos 
Godfrey 
Mohamed Gouda 
Albert G. Greensberg 
S Helal 
T Imielinski 
Subhash C Kak 
Abe Kandel 
Teuvo Kohomen 
Timo Koski 
Devendra Kumar 
Tosiyasu L Kunii 
Shahram Latifi 
Lin-shan Lee 
Mark Levene 
Jason Lin 
M i Lu 
Yanni  Manolopoulos 
Jorge Muruzabal 
Sham B Navathe 
Sigeru Ornatu 
C V Ramamoorthy 
Hari N. Reddy 
John R Rice 
Abdellah Salhi 
Frank S.Shih 
Harpreet Singh 
Stanley Y W. Su 
Abdullah Uz Tansel 
Kishor Trivedi 
Millist Vincent 
Benjamin Wah 
Z.A Wahab 
Jun Wang 
Patrick Wang 
Edward  K. Wong 
Lotfi A Zadeh

TOPICS: 1st ANNUAL COMPUTER THEORY  & INFORMATICS CONFERENCE

The conference will consist of both plenary sessions and contributory
sessions, focusing on topics of critical interest and the direction of
future research.  Example topics include, but are not limited to the
following:

o Computational Theory: Coding theory, automata, information theory, modern
algebraic theory, measure theory, probability and statistics, and numerical
methods

o Design of Algorithms: algorithmic complexity, theory of algorithms,
design, analyis and evaluation of efficient algorithms for engineering
applications (such as computer-aided design, computational fluid dynamics,
computer graphics, and virtual reality), combinatorics, scheduling theory,
discrete optimization, data compression, and approximation theory

o Software Design: Formal languages, theory and design of optimizing
compilers (especially those for parallel and supercomputers),
object-oriented programming, database theory and data organization,
software design methodology, program verification, and software
reliability.

o Computer systems and architectures: parallel and distributed computing
systems, high speed computer networks, theory and data orgarization,
software design methodology, program verification, and software
reliability.

o Evaluation methods and tools: Performance evaluation methods,
visualization tools, and simulation theory and methodology.

TOPICS: WORKSHOP ON MOBILE COMPUTING SYSTEMS

The workshop will focus on system support for mobile information and data
access, to recognize the role of mobile computer systems in today's
business and scientific communities. The workshop will be organized to
gather leading researchers and practitioners in the field. We shall focus
on issues related but not limited to: 

o Architectures for Mobile Computing Systems 
o Mobile ComputingTechnology 
o Wireless Communications 
o User lnterfaces in Palmtop Computing 
o Databases for Nomadic Computing 
o Transaction Models and Management 
o System Cornplexity, Integrity and Security
o Legal/Social/Health Issues  
o Operating System Support 
o Battery Management 
o Nomadic Applications
o Handheld Multimedia
o Personai Communication Networks

PROGRAM COMMITTEE: WORKSHOP ON MOBILE COMPUTING SYSTEMS 

R Alonso (Technology) 
B Badrinath (Query Processing) 
O Bukhres (Database Systems) 
P Chrysantis (Transactions) 
S Helal (Applications) 
T Imielinski (New Directions)

Note:
All Attendees must choose either Participation Plan A or Participation Plan B

PUBLICATIONS

The conference publishes two Proceedings on Summary; One entitled "Third
International Conference on Fuzzy Theory and Technology" and the other
entitled "First International Conference on Computer Theory and
Informatics, and First Workshop on Mobile Computing Systems." Both
proceedings will be made available on November 13,1994.

A summary shall not exceed 4 pages of 10-point font, doublecolumn,
single-spaced text, (1 page minimum) with figures and tables
included. Any summary exceeding 4 pages will be charged $100 per
additional page. Three copies of the summary are required by September
10, 1994. It is very important to mark "plan A" or "plan B" on your
manuscript. The conference will make the choice for you if you forget
to do so. Final version of the full length paper must be submitted by
November 14, 1994. Four (4) copies of the full length paper shall be
prepared according to the "Information for Authors" appearing at the
back cover of Information Sciences, an International Journal (Elsevier
Publishing Co.). A full paper shall not exceed 20 pages including
figures and tables. All full papers will be reviewed by experts in
their respective fields. Revised papers will be due on April 15, 1995.
Accepted papers will appear in the hard-covered proceeding (book) with
uniform typesetting to be published by a publisher (there will be two
books published this year, one for each plan) or Information Sciences
Journal (INS journal now has three publications: Informatics and
Computer Science, Intelligent Systems, Applications). All fully
registered conference attendees will receive a copy of proceeding
(summary) on November 14, 1994; a free one-year subscription (paid by
this conference) of Information Sciences Journal -
Applications. Lastly, the right to purchase either or both
hard-covered, deluxe, professional books at 1/2 price. The Title of
the books are "Advances in Fuzzy Theory & Technology, Volume lll",
"Advances in Computer Science and Informatics, Volume 1."

Lotfi A. Zadeh "Best Paper Award" FT&T 1994

All technical papers subrnitted to FT & T, 1994 are automatically qualified
as candidates for this award. The prize for this award is $2,500 plus hotel
accommodations (traveling expenses excluded) at FT & T, 1995. The date for
announcement of th best paper is May 30,1995. Oral presentation in person
at FT & T, 1994, is required and an acceptance speech at FT & T, 1995 is
also required. The evaluation committee for FT & T, 1994 consists of the
following 10 members:

Jack Aldridge 
B. Bouchon-Meunier 
George Klir l.R. Goodman 
John Mordeson 
Sujeet Shenoi 
H. Chris Tseng 
Frank Y. Shih 
Akira Nakamura 
Edward K. Wong 
I. B. Turksen.

The selection of the top ten best papers will be decided by conference
attendees and session chairs jointly.

EXHIBITIONS

JCIS '94 follows on last year's highly successful exhibits by some major
publishers, most publishers, lead by Elsevier, will return. Intelligent
Machines, Inc. will demonstrate their highly successful new software
"O'inca"-a FL-NN, Fuzzy-Neuro Design Framework. Dr. Ken W. White of Ithaca,
N.Y. will demonstrate his visual-sense systems. Virtus - a virtual reality
company based in Cary, North Carolina has committed to participate.
Negotiations are also underway with UNCVR research laboratory for its
participation. This conference intends to develop "virtual reality" as one
of the themes to benefit all attendees. Interested potential contributors
should contact Dr. Paul P. Wang or Dr. Rhett T. George.

Interested vendors should contact: Rhett George, E.E. Dept., Duke
University Telephone: (919) 660-5242 Fax: (919) 660-5293 rtg at ee.duke.edu

TRAVEL ARRANGEMENTS

The Travel Center of Durham, Inc. has been designated the official travel
provider. Special domestic fares have been arranged and The Travel Center
is prepared to book all flight travel. Domestic United States and Canada:1
-800-334-1 085 International FAX: (919) 687-0903

HOTEL RESERVATIONS Pinehurst Resort & Country Club Pinehurst, North
Carolina, U.S.A.

This is the conference site and lodging. Group Reservation Request designed
specifically for our conference. Very Special discount rates have been
agreed upon. Daily Rates for Hotel:

Single Occupancy - $122.00 
Double Occupancy - $91.00 per person 
Daily Rates for Manor Inn: Single Occupancy - $108.00 
Double Occupancy - $79.00/person 
(Rates are per person, per night and include accommodations, breakast and
dinner daily.)

Pinehurst Resort encompasses an elegant historic hotel (registered with
Historic Hotels of America") with the best in accommcations, gourmet dining
and modern meeting facilities. Our AAA Four Diamond and Mobil Four Star
resort offers a wide range of activities including seven championship golf
courses, tennis, waters sports, croquet and sport shooting.

Please contact: 
JACKIE HAYTER 
Associate Director of Sales 
Pinehurst Resort and Country Club 
Carolina Vista 
P.O. Box 4000 
Pinehurst, NC 28374-4000 
(919) 295-1339  - (919) 295-8484 
1-800-659-G0LF 

SPONSORS

Machine Intelligence and Fuzzy Logic Laboratory o Department of Electrical
Engineering, Duke University o Elsevier Science Inc. New York, N.Y.

PARTICIPANTS

IFSA, International Fuzzy Systems Association o Institute Of Information
science, Academia Sinica.

JCIS '94 REGISTRATION FEES & INFO.

                           Up to 9/15/94   After 9/15/94 
Full Registration            $275.00          $395.00 
Student Registration          $85.00          $160.00 
Tutorial w/ Conf_ Reg.       $150.00          $200.00 
Tutorial w/o Conf. Reg.      $300.00          $500.00 
Exhibit Booth Fee            $400.00          $500.00
One Day Fee 
(no pre-reg. discount)       $165.00 Full     $80.00 Student

Above fees applicable to both Plan A & Plan B

FULL CONFERENCE REGISTRATION: Includes admission to all sessions, exhibit
area, coffee, tea and soda A copy of conference proceedings (summany) at
conference and one year subscnption of Information Sciences - Applications,
An International Journal, published by Elsevier Publishing Co. In addition,
the right to purchase the hard-cover deluxe books at 1/2 price. Banquets
(November 14 & 15, 1994) are included through Hotel Registration. Tutorials
are not included

STUDENT CONFERENCE REGISTRATION. For full-time students only. A letter from
your department is required. You must present a current student I D with
picture. A copy of Conference Proceedings (Summary) is included. Admission
to all sessions, exhibit area, coffee, tea and soda. The nght to purchase
the hard-cver deluxe books at 1/2 price Free Subscription of INS Journal -
Applications is not included.

TUTORIALS REGISTRATION: Any person can register for the Tutorials A copy of
lecture notes for the oaurse registered is included. Coffee, tea and soda
is included The summary and tree subscription to the INS journal is,
however, not included. The right to purchase hard-cover deluxe books is
included.

VARIOUS CONFERENCE CONTACTS:

LOCAL INFORMATlON                                 
Rhett T. George Dept. of Electrical Engineering Box 90291, Duke University,
Durham, NC 27708-0291 e-mail: rtg at ee.duke.edu Tel. (919) 660-5228

TUTORIAL & CONFERENCE INFORMATION 
Paul P. Wang                          Kitahiro Kaneda 
e-mail: ppw at ee.duke.edu               e-mail: hiro at ee.duke.edu    
Tel. (919) 660-5271, 660-5259         Tel. (919) 660-5233 

Jerry C.Y. Tyan 
e-mail: ctyan at ee.duke.edu 
Tel. (919) 660-5233             
----------------------------------------------------------------

Neural Networks papers: Send summaries to

George M. Georgiou
Computer Science Department                  TEL: (909) 880-5332
California State University	             FAX: (909) 880-7004
5500 University Pkwy
San Bernardino, CA 92407, USA
georgiou at silicon.csci.csusb.edu

Deadline: September 10, 1994

TeX/Latex or postcript by email is fine.
--------------------------------------------------------------------