From wjf at eng.cam.ac.uk Wed Apr 1 03:58:02 1998 From: wjf at eng.cam.ac.uk (W J Fitzgerald) Date: Wed, 01 Apr 1998 08:58:02 +0000 Subject: BAYESIAN SIGNAL PROCESSING Message-ID: <117594.3100409882@wjfmac.eng.cam.ac.uk> Dear Friends, I hope this might be of interest to some of you. regards Bill Fitzgerald Isaac Newton Institute for Mathematical Sciences EC SUMMER SCHOOL BAYESIAN SIGNAL PROCESSING 19 - 31 July 1998 Organisers: WJ Fitzgerald (Cambridge), RL Smith (North Carolina), AT Walden (Imperial), PC Young (Lancaster) The main focus and thrust of this workshop will be that Bayesian methods provide a unifying methodology whereby different kinds of mathematical models may be examined within a common statistical framework. The workshop will bring together the statistical and computational expertise of leading statisticians and the modelling expertise of mathematicians and subject matter specialists, with the broad objective of developing new signal processing tools which make efficient use of modern computational resources while combining the most up-to-date research of both groups of specialists. Specific topics to be covered include: - Bayesian methods in general and numerical methods in particular; - Nonlinear and nonstationary time series estimation; - Forecasting and changepoint modelling; - Nonlinear signal processing in econometrics and financial time series; - Dynamical systems and statistics; - Environmental applications and spatial data analysis. The workshop will take place at the start of a six month programme on Nonlinear and Nonstationary Signal Processing to be held at the Isaac Newton Institute in Cambridge (July - Dec 1998), where it is hoped that many of the problems identified during the workshop will be studied in detail by the participants. Grants: The conference is supported by a grant from the European Community which will provide funding towards the registration, travel and subsistence costs of selected young (under 35 years) participants. Applications from women and anyone living in Greece, Ireland and Portugal and other less favoured regions of the European Community are particularly encouraged. Other limited funds exist for participants from outside the EC. Self-supporting applications of any age and nationality are welcome. Applications: The workshop will take place at the Newton Institute and accommodation for participants will be provided at Christ's College. The conference package costs ?650, which includes registration fees, accommodation, breakfast and evening meals plus lunch and refreshments during the days that lectures take place. Further Information and Application Forms: are available from the WWW at http://www.newton.cam.ac.uk/programs/nspw01.html Completed application forms should be sent to Heather Hughes at the Newton Institute, or via email to h.hughes at newton.cam.ac.uk The programme home page is at http://www.newton.cam.ac.uk/programs/nsp.html **Closing Date for the receipt of applications is 24 April 1998** ******************************************************************** * Dr. W.J.Fitzgerald, http://www-sigproc.eng.cam.ac.uk/~wjf * * Signal Processing Group, Cambridge University Engineering Dept. * * Cambridge CB2 1PZ, U.K. * * and/or * * Christ's College, * * Cambridge CB2 3BU * * * * tel: +44-(1223)-332719, fax: +44-(1223)-332662 * * email: wjf at eng.cam.ac.uk * ******************************************************************** From zemel at U.Arizona.EDU Thu Apr 2 18:40:26 1998 From: zemel at U.Arizona.EDU (Richard Zemel) Date: Thu, 2 Apr 1998 16:40:26 -0700 (MST) Subject: postdoctoral position Message-ID: POSTDOCTORAL RESEARCH ASSOCIATE UNIVERSITY OF ARIZONA Tucson, Arizona A postdoctoral position is available in the laboratory of Richard Zemel. It is expected that the applicant will collaborate in research projects in at least one of these areas: (1) theoretical models of neural coding; (2) studies of spatial and object attention; (3) motion processing in natural and artificial systems; (4) cue combination. For more details, see http://www.u.arizona.edu/~zemel. Numerous opportunities exist for interactions and collaborations with other researchers at the University of Arizona. The university has a strong interdisciplinary group in the area of Cognition and Neural Systems, including active research in: neural coding and spatial cognition (Carol Barnes, Bruce McNaughton, Lynn Nadel); object perception and attention (Mary Peterson); visual neurophysiology (Peter DeWeerd, Fraser Wilson); multimodal integration (Felice Bedford, Kerry Green) and others. For more information see http://w3.arizona.edu/~psych/cns.htm. Applicants should have a strong background and education in a quantitative discipline, such as physics, mathematics, statistics, or computer science. Knowledge of neuroscience or psychophysics is also desirable. Starting date is flexible, ranging from immediately to September. The position is available for 1-2 years depending on accomplishment. Salary is competitive. Please send a CV, a letter describing research interests and background, and at least 2 letters of recommendation by post or email to: Richard Zemel Department of Psychology University of Arizona Tucson, AZ 85721 Email: zemel at u.arizona.edu From becker at curie.psychology.mcmaster.ca Thu Apr 2 20:31:56 1998 From: becker at curie.psychology.mcmaster.ca (Sue Becker) Date: Thu, 2 Apr 1998 20:31:56 -0500 (EST) Subject: COMPUTATIONAL NEUROSCIENCE POSTDOC Message-ID: COMPUTATIONAL NEUROSCIENCE POSTDOCTORAL POSITION AVAILABLE Department of Psychology McMaster University A postdoctoral position is open in the Psychology Department of McMaster University, Hamilton, Ontario. A multidisciplinary approach will be taken to develop biologically plausible models of hippocampal and neocortical memory systems. Projects will include developing simulations of cortical-cortical and subcortical-cortical interactions during learning and information storage. In addition to our focus on processing in hippocampal-cortical systems, we are also investigating and modelling the role of cortico-thalamic back-projections. Research projects will be conducted in close collaboration with R. Racine, a neuroscientist, S. Becker, a computational modeller, and S. Haykin, an electrical engineer. The primary goal of this collaborative effort is to build powerful learning algorithms in neural networks which are based on rules suggested by both memory research and physiology research (e.g. LTP work). Racine's laboratory has recently provided the first demonstrations of LTP in the neocortex of the awake, freely-moving rat. The rules that apply to LTP induction in neocortical systems are quite different from those determined for the hippocampus. An OPTIONAL component of this postdoctoral position would be participation in further experimental investigations of neocortical LTP in either slice or in vivo preparations. This project is funded by a collaborative research grant from the Natural Sciences and Engineering Research Council of Canada to R. Racine, S. Haykin and S. Becker. Please send curriculum vitae, expression of interest, and the names and e-mail or phone numbers of three references to Ron Racine at racine at mcmail.cis.mcmaster.ca From jb at uran.informatik.uni-bonn.de Fri Apr 3 08:37:19 1998 From: jb at uran.informatik.uni-bonn.de (Joachim M. Buhmann) Date: Fri, 03 Apr 1998 15:37:19 +0200 Subject: Tech. Report on Unsupervised Learning Message-ID: <1.5.4.32.19980403133719.00665340@uran.informatik.uni-bonn.de> The following technical report on Unsupervised Learning is now available from our website http://www-dbv.informatik.uni-bonn.de/papers.html#NeuralNetworks Empirical Risk Approximation: An Induction Principle for Unsupervised Learning Joachim M. Buhmann Institute for Computer Science (III), University of Bonn Unsupervised learning algorithms are designed to extract structure from data without reference to explicit teacher information. The quality of the learned structure is determined by a cost function which guides the learning process. This paper proposes Empirical Risk Approximation as a new induction principle for unsupervised learning. The complexity of the unsupervised learning models are automatically controlled by the two conditions for learning: (i) the empirical risk of learning should uniformly converge towards the expected risk; (ii) the hypothesis class should retain a minimal variety for consistent inference. The maximal entropy principle with deterministic annealing as an efficient search strategy arises from the Empirical Risk Approximation principle as the optimal inference strategy for large learning problems. Parameter selection of learnable data structures is demonstrated for the case of K-means clustering. --------------------------------------------------------------------- Joachim M. Buhmann Institut fuer Informatik III Tel.(office) : +49 228 734 380 Universitaet Bonn Tel.(secret.): +49 228 734 292 Roemerstr. 164 Fax: +49 228 734 382 D-53117 Bonn email: jb at informatik.uni-bonn.de Fed. Rep. Germany jb at cs.bonn.edu http://www-dbv.informatik.uni-bonn.de --------------------------------------------------------------------- From wahba at stat.wisc.edu Fri Apr 3 17:24:19 1998 From: wahba at stat.wisc.edu (Grace Wahba) Date: Fri, 3 Apr 1998 16:24:19 -0600 (CST) Subject: Model Selection, RanTraceGACV Message-ID: <199804032224.QAA02784@hera.stat.wisc.edu> Im taking the liberty of noting that the following paper will be of interest to those interested in model selection: Jianming Ye, `On measuring anc correcting the effects of data mining and model selection', J. Amer. Statist. Assoc. 93, March 1998, pp 120- 131. ......... The following short abstract discusses the randomized trace technique for computing the GACV (Generalized Approximate Cross Validation function) for Bernoulli Data which was proposed in Xiang and Wahba, Statistica Sinica 1996, 675-692. Xiang, D. and Wahba, G. `Approximate Smoothing Spline Methods for Large Data Sets in the Binary Case', TR 982. http://www.stat.wisc.edu/~wahba -> TRLIST From baluja at jprc.com Mon Apr 6 15:18:01 1998 From: baluja at jprc.com (Shumeet Baluja) Date: Mon, 6 Apr 1998 15:18:01 -0400 Subject: paper available: Fast Probabilistic Modeling for Combinatorial Optimization Message-ID: <199804061918.PAA01197@india.jprc.com> Fast Probabilistic Modeling for Combinatorial Optimization ----------------------------------------------------------- Shumeet Baluja Justsystem Pittsburgh Research Center & Carnegie Mellon University Scott Davies Carnegie Mellon University Abstract Probabilistic models have recently been utilized for the optimization of large combinatorial search problems. However, complex probabilistic models that attempt to capture inter-parameter dependencies can have prohibitive computational costs. The algorithm presented in this paper, termed COMIT, provides a method for using probabilistic models in conjunction with fast search techniques. We show how COMIT can be used with two very different fast search algorithms: hillclimbing and Population-based incremental learning (PBIL). The resulting algorithms maintain many of the benefits of probabilistic modeling, with far less computational expense. Extensive empirical results are provided; COMIT has been successfully applied to jobshop scheduling, traveling salesman, and knapsack problems. This paper also presents a review of probabilistic modeling for combinatorial optimization. This paper is an extension of the earlier work reported in: Combining Multiple Optimization Runs with Optimal Dependency Trees by: S. Baluja & S. Davies, June 1997. Available from: http://www.cs.cmu.edu/~baluja Questions and comments are welcome. Shumeet Baluja From aapo at myelin.hut.fi Tue Apr 7 12:32:32 1998 From: aapo at myelin.hut.fi (Aapo Hyvarinen) Date: Tue, 7 Apr 1998 19:32:32 +0300 Subject: FastICA package for MATLAB Message-ID: <199804071632.TAA26001@myelin.hut.fi> FastICA, a new MATLAB package for independent component analysis, is now available at: http://www.cis.hut.fi/projects/ica/fastica/ FastICA is a public-domain package that implements the fast fixed-point algorithm for ICA, and features an easy-to-use graphical user interface. The fixed-point algorithm is a computationally highly efficient method for ICA: in independent experiments it has been found to be 10-100 times faster than conventional gradient descent methods for ICA. Another advantage of the fixed-point algorithm is that it can be used to perform projection pursuit, estimating the independent components one-by-one. Aapo Hyvarinen on behalf of the FastICA Team at the Helsinki University of Technology fastica at mail.cis.hut.fi From marco at idsia.ch Tue Apr 7 08:58:29 1998 From: marco at idsia.ch (Marco Wiering) Date: Tue, 7 Apr 1998 14:58:29 +0200 Subject: Paper announcement !!!!!!! Message-ID: <199804071258.OAA03950@ruebe.idsia.ch> Fast Online Q(lambda) Marco Wiering Juergen Schmidhuber To appear in the Machine Learning Journal Q(lambda)-learning uses TD(lambda)-methods to accelerate Q-learning. The update complexity of previous online Q(lambda) implementations based on lookup-tables is bounded by the size of the state/action space. Our faster algorithm's update complexity is bounded by the number of actions. The method is based on the observation that Q-value updates may be postponed until they are needed. Also to be presented at the 10th European Conference On Machine Learning (ECML'98), Chemnitz (Germany), April 21-24 1998. FTP-host: ftp.idsia.ch FTP-files: /pub/marco/fast_q.ps.gz /pub/marco/ecml_q.ps.gz WWW: http://www.idsia.ch/~marco/publications.html http://www.idsia.ch/~juergen/onlinepub.html Marco & Juergen IDSIA, Switzerland From todd at cs.ua.edu Thu Apr 9 16:41:16 1998 From: todd at cs.ua.edu (Todd Peterson) Date: Thu, 9 Apr 1998 15:41:16 -0500 (CDT) Subject: Paper Available Message-ID: <199804092041.PAA02777@todd.cs.ua.edu> Paper available: An RBF Network Alternative for a Hybrid Architecture Todd Peterson, Ron Sun Department of Computer Science The University of Alabama Tuscaloosa, AL 35487 EMAIL: todd,rsun at cs.ua.edu ABSTRACT Although a previous model CLARION has shown some measure of success in sequential decision making tasks by utilizing a hybrid architecture that uses both procedural and declarative learning, it had some problems resulting from the use of backpropagation networks. CLARION-RBF is a more parsimonious architecture that remedies some of the problems exhibited in CLARION, by utilizing RBF Networks. CLARION-RBF is capable of learning reactive procedures as well as having high level symbolic knowledge extracted and applied. To appear in IJCNN, Anchorage, AK, May 4-9, 1998. The Postscript version is accessible through: http://cs.ua.edu/~rsun/tp.rbf.ps or from: http://cs.ua.edu/~todd/ijc.ps From sknerr at ireste.fr Fri Apr 10 12:06:44 1998 From: sknerr at ireste.fr (stefan knerr) Date: Fri, 10 Apr 1998 18:06:44 +0200 Subject: PhD Thesis Message-ID: <352E4394.3218@ireste.fr> PhD Studentship in Document Image Processing Department of Image and Video Processing University Nantes - IRESTE, France DOCUMENT IMAGE MODELING AND SEGMENTATION USING STATISTICAL 2-D APPROACHES Early processing stages of document analysis systems such as the localization and segmentation of text fields, logos, graphics, lines, etc. usually rely on heuristic methods or must be defined by hand. The goal of this thesis is to elaborate a principled approach to these problems based on statistical approaches such as Markov Random Fields and 2-D Hidden Markov Models. This is an exciting opportunity to work on a real world problem within an environment which includes an industrial partner as well as a University research lab. Several possibilities for financing a 3 year PhD work exist depending on the qualification and on the nationality of the candidate. Interested candidates should send a CV, related publications or reports, names and addresses of one or two persons (professor, lecturer) or companies who will recommend you to: Stefan Knerr (sknerr at ireste.fr) or Christian Viard-Gaudin (cviard at ireste.fr) IRESTE Rue Christian Pauc, La Chantrerie, BP 60601 44306 Nantes Cedex 3 France Applications should arrive before Mai 31, 1998. -- Stefan Knerr Laboratoire SEI, IRESTE Rue Christian Pauc, La Chantrerie, BP 60601 44306 Nantes Cedex 3, France Tel. +33 2 40683036 Fax. +33 2 40683066 sknerr at ireste.fr From ingber at ingber.com Mon Apr 13 13:54:19 1998 From: ingber at ingber.com (Lester Ingber) Date: Mon, 13 Apr 1998 12:54:19 -0500 Subject: Paper: Volatility Volatility of Financial Markets Message-ID: <19980413125419.A3006@ingber.com> The paper markets98_vol.ps.Z [120K] is available at my InterNet archive: %A L. Ingber %A J.K. Wilson %T Volatility of volatility of financial markets %R DRW-98-1-VVFM %I DRW Investments LLC %C Chicago, IL %D 1998 %O URL http://www.ingber.com/markets98_vol.ps.Z We present empirical evidence for considering volatility of Eurodollar futures as a stochastic process, requiring a generalization of the standard Black-Scholes (BS) model which treats volatility as a constant. We use a previous development of a statistical mechanics of financial markets (SMFM) to model these issues. ======================================================================== Instructions for Retrieval of Code and Reprints Interactively Via WWW The archive can be accessed via WWW path http://www.ingber.com/ http://www.alumni.caltech.edu/~ingber/ where the last address is a mirror homepage for the full archive. Interactively Via Anonymous FTP Code and reprints can be retrieved via anonymous ftp from ftp.ingber.com. Interactively [brackets signify machine prompts]: [your_machine%] ftp ftp.ingber.com [Name (...):] anonymous [Password:] your_e-mail_address [ftp>] binary [ftp>] ls [ftp>] get file_of_interest [ftp>] quit The 00index file contains an index of the other files. Files have the same WWW and FTP paths under the main / directory; e.g., http://www.ingber.com/MISC.DIR/00index_misc and ftp://ftp.ingber.com/MISC.DIR/00index_misc reference the same file. Electronic Mail If you do not have WWW or FTP access, get the Guide to Offline Internet Access, returned by sending an e-mail to mail-server at rtfm.mit.edu with only the words send usenet/news.answers/internet-services/access-via-email in the body of the message. The guide gives information on using e-mail to access just about all InterNet information and documents. Additional Information Limited help assisting people with queries on my codes and papers is available only by electronic mail correspondence. Sorry, I cannot mail out hardcopies of code or papers. Lester ======================================================================== -- /* Lester Ingber Lester Ingber Research * * ingber at ingber.com http://www.ingber.com/ ftp.ingber.com * * ingber at alumni.caltech.edu http://www.alumni.caltech.edu/~ingber/ * * PO Box 06440 Wacker Dr PO - Sears Tower Chicago, IL 60606-0440 */ From aperez at lslsun.epfl.ch Tue Apr 14 04:12:41 1998 From: aperez at lslsun.epfl.ch (Andres Perez-Uribe) Date: Tue, 14 Apr 1998 10:12:41 +0200 Subject: New Book: Bio-Inspired Computing Machines Message-ID: <35331A79.E99B3A3D@lslsun.epfl.ch> Dear Connectionist, This is to announce a new book entitled: "Bio-Inspired Computing Machines Toward Novel Computational Architectures" Daniel Mange and Marco Tomassini (Eds.) Presses polytechniques et universitaires romande, Lausanne, Switzerland http://lslwww.epfl.ch/pages/publications/books/1998_1/ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Originality: This book is unique for the following reasons: - It follows a unified approach to bio-inspiration based on the so-called POE model: phylogeny (evolution of species), ontogeny (development of individual organisms), and epigenesis (life-time learning). - It is largely self-contained, with an introduction to both biological mechanisms (POE) and digital hardware (digital systems, cellular automata). - It is mainly applied to computer hardware design. - It is largely self-contained, with an introduction to both biological mechanisms (POE) and digital hardware (digital systems, cellular automata). - It is mainly applied to computer hardware design. BACK-COVER TEXT This volume, written by experts in the field, gives a modern, rigorous and unified presentation of the application of biological concepts to the design of novel computing machines and algorithms. While science has as its fundamental goal the understanding of Nature, the engineering disciplines attempt to use this knowledge to the ultimate benefit of Mankind. Over the past few decades this gap has narrowed to some extent. A growing group of scientists has begun engineering artificial worlds to test and probe their theories, while engineers have turned to Nature, seeking inspiration in its workings to construct novel systems. The organization of living beings is a powerful source of ideas for computer scientists and engineers. This book studies the construction of machines and algorithms based on natural processes: biological evolution, which gives rise to genetic algorithms, cellular development, which leads to self-replicating and self-repairing machines, and the nervous system in living beings, which serves as the underlying motivation for artificial learning systems, such as neural networks. PUBLIC Undergraduate and graduate students, researchers, engineers, computer scientists, and communication specialists. TABLE OF CONTENTS Preface 1 An Introduction to Bio-Inspired Machines 2 An Introduction to Digital Systems 3 An Introduction to Cellular Automata 4 Evolutionary Algorithms and their Applications 5 Programming Cellular Machines by Cellular Programming 6 Multiplexer-Based Cells 7 Demultiplexer-Based Cells 8 Binary Decision Machine-Based Cells 9 Self-Repairing Molecules and Cells 10 L-hardware: Modeling and Implementing Cellular Development using L-systems 11 Artificial Neural Networks: Algorithms and Hardware Implementation 12 Evolution and Learning in Autonomous Robotic Agents Bibliography Index -- Andres PEREZ-URIBE Logic Systems Laboratory Computer Science Department Swiss Federal Institute of Technology-Lausanne 1015 Lausanne, Switzerland Email: aperez at lslsun.epfl.ch http://lslwww.epfl.ch/~aperez Tel: +41-21-693-2652 Fax: +41-21-693 3705 From harnad at coglit.soton.ac.uk Tue Apr 14 09:39:12 1998 From: harnad at coglit.soton.ac.uk (S.Harnad) Date: Tue, 14 Apr 1998 14:39:12 +0100 (BST) Subject: Connectionist Explanation: PSYC Call for Commentators (614 lines) Message-ID: <199804141339.OAA15840@amnesia.psy.soton.ac.uk> Green: CONNECTIONIST EXPLANATION The target article below has just appeared in PSYCOLOQUY, a refereed journal of Open Peer Commentary sponsored by the American Psychological Association. Qualified professional biobehavioral, neural or cognitive scientists are hereby invited to submit Open Peer Commentary on this article. Please email for Instructions if you are not familiar with PSYCOLOQUY format and acceptance criteria (all submissions are refereed). To submit articles and commentaries or seek information: EMAIL: psyc at pucc.princteton.edu URL: http://www.princeton.edu/~harnad/psyc.html http://www.cogsci.soton.ac.uk/psyc AUTHOR'S RATIONALE FOR SOLICITING COMMENTARY: My reason for soliciting commentary is quite straightforward. Connectionist models of cognition are quite common in the psychological literature these days, but there is very little discussion of the exact role they are thought to play in science. If they are indeed theories, in the traditional sense, then some explanation of the ways in which they seem to depart from traditional theories is needed. If they are not traditional theories, then a clear description of what they are, and an account of why we should pay attention to them, is needed. Such a discussion should take place among connectionists, philosophers of science and mind, and psychologists. Psycoloquy seems like an ideal vehicle for such a discussion. ----------------------------------------------------------------------- psycoloquy.98.9.04.connectionist-explanation.1.green Tue 14 Mar 1998 ISSN 1055-0143 (23 paragraphs, 16 references, 4 notes, 584 lines) PSYCOLOQUY is sponsored by the American Psychological Association (APA) Copyright 1998 Christopher D. Green ARE CONNECTIONIST MODELS THEORIES OF COGNITION? Christopher D. Green Department of Psychology York University North York, Ontario M3J 1P3 CANADA christo at yorku.ca http://www.yorku.ca/faculty/academic/christo ABSTRACT: This paper explores the question of whether connectionist models of cognition should be considered to be scientific theories of the cognitive domain. It is argued that in traditional scientific theories, there is a fairly close connection between the theoretical (unobservable) entities postulated and the empirical observations accounted for. In connectionist models, however, hundreds of theoretical terms are postulated -- viz., nodes and connections -- that are far removed from the observable phenomena. As a result, many of the features of any given connectionist model are relatively optional. This leads to the question of what, exactly, is learned about a cognitive domain modelled by a connectionist network. KEYWORDS: artificial intelligence, cognition, computer modelling, connectionism, epistemology, explanation, methodology, neural nets, philosophy of science, theory. 1. Connectionist models of cognition are all the rage now. It is not clear, however, in what sense such models are to be considered THEORIES of cognition. This may be problematic, for if connectionist models are NOT to be considered THEORIES of cognition, in the traditional scientific sense of the word, then the question arises as to what exactly they are, and why we should pay attention to them? If, on the other hand, they are to be regarded as scientific theories it should be possible to explicate precisely in what sense this is true, and to show how they fulfill the functions we normally associate with theories. In this paper, I begin by examining the question of what it is to be a scientific theory. Second, I describe in precisely what sense traditional computational models of cognition can be said to perform this role. Third, I examine whether or not connectionist models can be said to do the same. My conclusion is that connectionist models could, under a certain interpretation of what it is they model, be considered to be theories, but that this interpretation is likely to be unacceptable to many connectionists. 2. A typical complex scientific theory contains both empirical and theoretical terms. The empirical terms refer to observable entities. The theoretical terms refer to unobservable entities that improve the predictive power of the theory as a whole. The exact ontological status of objects referred to by theoretical terms is a matter of some debate. Realists believe them to be actual objects that resist direct observation for one reason or another. Instrumentalists consider them to be mere "convenient fictions" that earn their scientific keep merely by the predictive accuracy they lend to the theory. I think it is fair to say that the vast majority of research psychologists are realists about the theoretical terms they use, though they are, in the main, unreflective realists who have never seriously considered alternative possibilities. 3. Let us begin with a relatively uncontroversial theory from outside psychology -- Mendelian genetics. In the Mendelian scheme, entities called "genes" were said to be responsible for the propagation of traits from one generation of organisms to another. Mendel was unable to observe anything corresponding to "genes," but their invocation made it possible for him to predict correctly the proportions in which succeeding generations of organisms would express a variety of traits. As such, the gene is a classic example of a theoretical entity. For present purposes, it is important to note that each such theoretical gene, though unobservable, was hypothesized to correspond to an individual trait. That is, in addition to the predictive value each theoretical gene provided, each also justified its existence by being responsible for a particular phenomenon. There were no genes in the system that were not directly tied to the expression of a trait. Although some genes were said not to be expressed in the phenotype (viz., recessive genes in heterozygous individuals), all were said to be directly involved in the calculation of the expression of a specific trait. That is to say, their inclusion in the theory was justified in part by the SPECIFICITY of the role they were said to play. It is worth noting that the actual existence of genes remained controversial until the discovery of their molecular basis -- viz., DNA -- and our understanding of them changed considerably with that discovery. 4. Now consider, as a psychological example of theoretical entities, the model of memory proposed by Atkinson and Shiffrin (1971). It is a classic "box-and-arrow" theory. Information is fed from the sensory register into a holding space called Short Term Store (STS). If continuously rehearsed, a limited number of items can be stored there indefinitely. If the number of items exceeds the capacity of the store, some are lost. If rehearsal continues for an unspecified duration, it is claimed that some or all of these items are transferred to another holding space called Long Term Store (LTS). The capacity of LTS is effectively unlimited, and items in LTS need not be continuously rehearsed, but are said to be kept in storage effectively permanently. STS and LTS are, like genes, theoretical entities. They cannot be directly observed, but their postulation enables the psychologist to predict correctly a number of memory phenomena. In each such phenomenon, the activity of each store is carefully specified. The precision of this specification seems to be at least part of the reason that scientists are willing to accept them. Indeed, many experiments aimed at confirming their existence are explicitly designed to block, or interfere with the hypothesized activity of one in order to demonstrate the features of the "pure" activity of the other. Whether or not this could be successfully accomplished was once a dominant question in memory theory. The issue of short term memory EFFECTS being "contaminated" by the uncontrollable and unwanted activity of LTS occupied many experiments of the 1960s and 1970s. 5. Over the last 30 years the Atkinson and Shiffrin model has been elaborated and refined. As a result, the number of memory systems hypothesized to exist has grown tremendously. Baddeley (1992), for instance, has developed STS into a series of slave systems responsible for information entering memory from the various sense modalities (e.g., the phonological loop, the visuospatial sketchpad), the activities of which are coordinated by a central executive. Tulving (1985), on the other hand, has divided LTS into four hierarchically arranged systems responsible for episodic memory (for personal events), semantic memory (for general information), procedural memory (for skills) and implicit memory (for priming). In order to establish the existence of each of these many theoretical entities, thousands of experiments have been performed, aimed at revealing the independent of activity of one or another by attempts to block the activity of the others. 6. Once again, the question of whether the activity of a single memory system can be studied in isolation has called into question the very existence of that system. For over a decade, now, the elucidation of implicit memory phenomena has been a major issue in memory theory (Schacter, 1987, 1992; Roediger, 1990; Roediger & McDermott, 1993;). In the typical implicit memory experiment [1], subjects study a list of items (both words and pictures have been used) by processing them briefly. This can be as simple as reading the word or naming the object, or it can be more involved, such as deciding whether the items belongs to a certain class of items (e.g., is a car a kind vehicle?) or decomposing it in to parts (e.g., counting the number of letters in words, or counting the edges or corners in pictured items). The subjects then take part in a memory test, although they are not told that it is a memory test, and it could indeed be performed without having studied the material. In this test, they see a new list of items, some of which, unbeknownst to them, are the same as (or closely related to) the items they have studied. Such tests are sometimes puzzles of various sorts (e.g., completing incomplete words or identifying the items in incomplete pictures). Sometimes they are as simple as deciding whether the items are true words (as opposed to pronounceable non-words such as BLICK) or possible objects. People perform reliably better on these tasks when the items in question are ones that were on the study list (or closely related to items on the study list) than when the items are new. Upon post-experimental debriefing, however, they are often unable to say which items they had studied before and which they had not. This is the classic implicit memory effect. 7. Recently, however, it has been argued (Roediger & McDermott, 1993) that explicit memory may be "contaminating" the hypothesized effect of the implicit memory system. The degree of this contamination is not clear, but it is possible, in principle (though unlikely), that ALL implicit memory phenomena are the result of covert explicit memory. The evidence for this (Jacoby, 1991) comes from comparing the behavior of a typical implicit memory group with that of a control group that goes through the same procedure but is told EXPLICITLY that the answers to some of the test problems are items they studied before. The outcome is that these subjects do almost as well as the experimental subjects, thus calling into question the "implicitness" of the traditional subjects' memories. As a result, many have begun question the very existence of the implicit memory system. Many psychologists argue that the implicit memory effects are the result of a certain kind of processing of a more general memory system, not the autonomous activity of a distinct system of its own. 8. With the entry of computer models into psychology, the theories have become even more complex, using dozens of theoretical entities. A recent version of Chomskyan linguistic theory, for instance, postulates more than two dozen rules that are said to control the building and interpretation of grammatical sentences (see e.g., Berwick, 1985). But even here the empirical data must bear fairly directly on each theoretical entity. None of these rules is without specific predicted effects. Each of the rules performs a certain function without which the construction and interpretation of grammatical sentences could not proceed correctly. For example, RULE ATTACH-VP, sensibly enough, attaches verb phrases to sentences; RULE ATTACH-NOUN similarly attaches nouns to noun phrases; and so forth. Part of what justifies the inclusion in the theory of terms referring to each of these entities is the fact that they are explicitly connected to specific empirical phenomena. 9. In each of the models I have described so far, each theoretical entity represents something in particular, even if that something is itself theoretical. The existence and properties of the entities represented are supported by empirical evidence relating specifically to that entity. In a typical connectionist model, however, there are dozens, sometimes hundreds, of simple units, bound together by hundreds, sometimes thousands, of connections. Neither the units nor the connections represent anything known to exist in the cognitive domain the network is being used to model. Similarly, the rules that govern how the activity of one unit will affect the activity of other units to which it is connected are extremely simple, and not obviously related to the domain that the network is being used to model. Ditto for the rules that govern how the weights on the connections between units are to be changed. In particular, the units of the network are not thought to represent particular propositional attitudes (i.e., beliefs, desires, etc.) or the terms or concepts that might be thought to underlie them. This is all considered a distinct advantage among connectionists. Neither the units nor the connections correspond to anything in the way that variables and rules did in traditional computational models of cognition. Representations, to the degree that they are admitted at all, are said to be distributed across the activities of the units as a group. Any representation-level rules that the model is said to use are likewise distributed across the weights of all of the connections in the network. This gives connectionist networks their characteristic flexibility: they are able to learn in a wide variety of cognitive domains, to generalize their knowledge easily to new cases, to continue working reasonably well despite incomplete input or even moderate damage to their internal structure, etc. The only real question is whether they are, indeed, TOO flexible to be good theories. Or whether, by contrast, there are heretofore unrecognized features of good theories of which connectionist models can apprise us. 10. Each of the units, connections, and rules in a connectionist network is a theoretical entity. Each name referring to it in a description of the network is a theoretical term in the theory of cognition that it embodies [2]. With the previously described theories, it was evident that each theoretical entity had a specific job to do. If it were removed, not only would the performance of the model as a whole suffer, but it would suffer in predictable ways, viz., the particular feature of the model's performance for which the theoretical entity in question was responsible -- i.e., that which it represented -- would no longer obtain. The units and connections in a connectionist net -- precisely in virtue of the distributed nature of their activity -- need not bear any such relation to the various activities of the model. Although this seems to increase the model's overall efficiency, it also seems to undermine the justification for each of the units and connections in the network. To put things even more plainly, if one were to ask, say, of Berwick's (1985) symbolic model of grammar, "What is the justification for postulating RULE ATTACH-NOUN?" the answer would be quite straightforward: "Because without it nouns would not be attached to noun phrases and the resulting outputs would be ungrammatical." The answer to the parallel question with respect to the a connectionist network -- viz., "What is the justification for postulating (say) unit 123 in this network?" -- is not so straightforward. Precisely because connectionist networks are so flexible, the right answer is probably something like, "No reason in particular. The network would probably perform just as well without it" [3]. 11. If this is true, we are led to an even more pressing question: exactly what is it that we can actually be said to KNOW about a given cognitive process once we have modelled it with a connectionist network? In the case of, say, the Atkinson and Shiffrin model of memory, we can say that we have confirmation of the idea that there are at least two forms of memory store -- short and long term -- and this confirmation amounts to a justification of sorts for their postulation. Are we similarly to say that a particular connectionist model with, say, 326 units that correctly predicts activity in a given cognitive domain confirms the idea that there are exactly 326 units governing that activity? This seems ridiculous -- indeed almost meaningless. Aside from the obvious fact that we don't know what the "units" are units OF, we might well have gotten just as good results with 325, or 327 units, or indeed with 300 or 350 units. Since none of the units correspond to ANY particular aspect of the performance of the network, there is no particular justification for any one of them. Some might argue that the theory instantiated by the network is not meant to be read at this level of precision -- that it is not the number of units, specifically, that is being put forward for test, but only a network with a certain general sort of architecture and certain sorts of activation and learning rules. This seems simply too weak a claim to be of much scientific value. As Popper told us, scientists should put forward "bold conjectures" for test. The degree to which the hypothesis is subject to refutation by the test is the degree to which it is scientifically important. Even without accepting Popper's strong stand on the unique status of refutation in scientific work, this much remains clear: To back away from the details of one's theory -- to shield them from the possibility of refutation -- is to make one's theory scientifically less significant. Surely this is not a move connectionist researchers want to make in the long run. 12. It might be argued that the mapping of particular theoretical terms on to particular aspects of the behavior being modelled is unnecessary; it may just be an historical accident, primarily the result of our not being able to keep simultaneous control of thousands of theoretical terms until the advent of computers. Perhaps surprisingly, Carl Hempel seems to have presaged this possibility in his classic essay, Fundamentals of Concept Formation in Empirical Science: "A scientific theory might ... be likened to a complex spatial network: Its terms are represented by knots, while the threads connecting the latter correspond, in part, to the definitions and, in part, to the fundamental and derivative hypotheses included in the theory. The whole system floats, as it were, above the plane of observation and is anchored to it by rules of interpretation. These might be viewed as strings which are not part of the network but link to certain points of the latter with specific places in the plane of observation. By virtue of those interpretive connections, the network can function as a scientific theory: From certain observational data, we may ascend, via an interpretive string, to some point in the theoretical network, thence proceed, via definitions and hypotheses, to other points, from which another interpretive string permits a descent to the plane of observation." (Hempel, 1952, p. 36) 13. Now, it is by no means clear that Hempel had in mind here that there might be literally thousands of "knots in the network" between those few that are connected to the "plane of observation," but by the same token there is nothing in the passage that seems to definitely preclude the possibility either. 14. The real question seems to be about what one can really be said to have learned about the phenomenon of interest if one's model of that phenomenon contains far more terms that are not tied down to the "empirical plane," so to speak, than it does entities that are. Consider the following analogy: suppose that an historian wants to understand the events that lead up to political revolutions, so he tries to simulate several revolutions and a variety of other less successful political uprisings with a connectionist network. The input units encode data on, say, the state of the economy in the years prior to the uprising, the morale of the population, the kinds of political ideas popular at the time, and a host of other important socio- political variables. The output units encode various possible outcomes: revolution, uprising forcing significant political change, uprising diffused by superficial political concessions, uprising put down by force, etc. Among the input and output units, let us say that the historian places exactly 72 units which, he says, encode "a distributed representation of the socio-political situation of the time." His simulation runs beautifully. Indeed, let us say that because he has learned the latest techniques of recurrent networks, he is actually able to simulate events in the order in which they took place over several years either side of each uprising. 15. What has he learned about revolution? That there must have been (even approximately) 72 units involved? Certainly not. If the "hidden" units corresponded to something in particular -- say, to political leaders, or parties, or derivative socio-political variables -- that is, if the network had been SYMBOLIC, then perhaps he would have a case. Instead, he must simply repeat the mantra that they constitute "a distributed representation of the situation," and that the network is likely a close approximation to the situation because it plausibly simulates so many different variants of it. 16. It must be concluded that he has not learned very much about revolution at all. The simple fact of having a working "simulation" seems to mean little. It is only if one can interpret the INTERNAL ACTIVITY of the simulation that the simulation increases our knowledge; i.e., it is only then that the simulation is to be considered a scientific THEORY worthy of consideration. 17. Some might find this analogy invalid because of the widely recognized problems with studying history with the methods of science. My own opinion is that this is a non sequitur; but rather than arguing the point let us turn to a less controversial case. Assume for the moment that some aspiring amateur physicist, blithely unaware of the work of Galileo and Newton, gets the idea that the way to study the dynamics of balls rolling down inclined planes is to simulate their movements with a connectionist network. He sets up the net with inputs corresponding to variables such as the mass and volume of the ball, the length and angle of the plane, etc. Perhaps, not really knowing what he is after, he adds in some interesting variations such as ellipsoidal balls and curved surfaces, and includes the pertinent features of these in his encoding scheme. The activity of the output unit represents simply the time it takes the ball to complete its descent down the surface. He throws in a handful of hidden units, say 5, and runs the simulation. Eventually the network is able to predict closely how long it will take a certain ball to run down a certain surface, and it is able to generalize its knowledge to new instances on which it was not trained. If asked what the hidden units represent, the young physicist says, "the individual units represent nothing in particular; just a distributed representation of the physical situation as a whole." What has he learned? Not much, it would seem. Certainly not what was learned in the explanation of these kinds of phenomena in the theories of Galileo and Newton, in which the theoretical entities clearly REFER to relatively uncontroversial aspects of the world (e.g., distance, duration, size). 18. One way we cognitive scientists might try to avoid the fate of our hypothetical connectionist historian and physicist is to claim that connectionist units DO correspond to something closely related to the cognitive domain; viz., the neurons of the brain. Whether this is to be considered an analogy or an actual literal claim is often left vague by those who suggest it. Most connectionists seem wary of proclaiming too boldly that their networks model the actual activity of the brain. McClelland, Rumelhart, and Hinton (1986), for instance, say that connectionist models "seem much more closely tied to the physiology of the brain than other information-processing models" (p. 10), but then they retreat to saying that their "physiological plausibility and neural inspiration...are not the primary bases of their appeal to us" (p. 11). Smolensky (1988), after having examined a number of possible mappings, writes that "given the difficulty of precisely stating the neural counterpart of components of subsymbolic [i.e., connectionist] models, and given the very significant number of misses, even in the very general properties considered..., it seems advisable to keep the question open" (p. 9). Only with this caveat in place does he then go on to claim that "there seems no denying, however, that the subconceptual [i.e., connectionist] level is SIGNFICANTLY CLOSER [emphasis added] to the neural level than is the conceptual [i.e., symbolic] level" (p. 9). Precisely what metric he is using to measure the "closeness" of various theoretical approaches to the neural level of description is left unexplicated. 19. The general aversion to making very strong claims about the relation between connectionist models and brain is not without good reason. Crick and Asanuma (1986) describe five properties that the units of connectionist networks typically have that are rarely or never seen in neurons, and two further properties of neurons that are rarely found in the units of connectionist networks. Perhaps most important of these is the fact that the success of connectionist models seems to DEPEND upon the fact that any given unit can send excitatory impulses to some units and inhibitory impulses to others. No neuron in the mammalian brain is known to do this (though "dual-action" neurons have been found in the abdominal ganglion of Aplysia; see Levitan & Kaczmarek, 1991, pp. 196-197). Although it is certainly possible that dual-action neurons will be found in the human brain, the vast majority of cells do not seem to have this property, whereas the vast majority of units in connectionist networks typically do. Even as strong a promoter of connectionism as Paul Churchland (1990, p. 221) has recognized this as a major hurdle to be overcome if connectionist nets are to be taken seriously as models of brain activity. What is more, despite some obvious but possibly superficial similarities between the structure of connectionist units and the structure of neurons, there is currently little hard evidence that any SPECIFIC aspect of cognition is instantiated in the brain by neurons arranged in any SPECIFIED connectionist configuration. 20. It would accordingly appear that at present the only way of interpreting connectionist networks as serious candidates for theories of cognition, would be as literal models of the brain activity that underpins cognition. This means, if Crick and Asanuma are right in their critique, that connectionists should start restricting themselves to units, connections, and rules that use all and only principles that are known to be true of neurons. Other interpretations of connectionist networks may be possible in principle, but at this point none seem to have appeared on the intellectual horizon [4]. Without such an interpretation, connectionist modelers are left more or less in the position of out hypothetical connectionist historian. Even a simulation that is successful in terms of transforming certain inputs into the "right" outputs does not tell us much about the cognitive process it is simulating unless there is a plausible interpretation of its inner workings. All the researcher can claim is that the success of the simulation confirms that SOME connectionist architecture is involved, and perhaps something very general about the nature of that architecture (e.g., that it is self-organizing, recurrent, etc.). There is little or no confirmation of the specific features of the network because so much of it is OPTIONAL. 21. Now, it might be argued that this situation is no different from that of early atomic theory in physics. Visible bits of matter and their interactions with other bits of matter were explained by the postulation of not just thousands, but millions upon millions of theoretical entities of mostly unknown character -- viz., atoms. This, the argument would continue, is not so different from the situation in connectionism. After all, as Lakatos (1970) taught us, new research programs need a grace period in the beginning to get themselves established. Although I don't have a demonstrative argument against this line of thought, I think it has relatively little merit. We know pretty well what atoms are, and where we would find them, were we able to achieve the required optical resolution. Put very bluntly, if you simply look closer and closer and closer at a material object, you'll eventually see the atoms. Atoms are, at least in that sense, perfectly ordinary material objects themselves. Although they constitute an extension of our normal ontological categories, they do not REPLACE an old well-understood category with a new ill-understood one.[5] 22. By contrast, the units of connectionist networks (unless identified with neurons, or other bits of neural material) are quite different. They are not a REDUCTION of mental concepts, and as such give us no obvious path to follow to get from the "high level" of behavior and cognition to the "low level" of units and connections. That it is not a REDUCTIVE position is in fact often cited as a STRENGTH of connectionism but, if I am right, it is also the primary source of the ontological problems that have been discussed here. 23. To conclude, it is important to note that I am not arguing that connectionist networks must give way to symbolic networks because cognition is inherently symbolic (see, e.g., Fodor & Pylyshyn, 1988). That is an entirely independent question. What I am suggesting, however, is that the apparent success of connectionism in domains where symbolic models typically fail may be due as much to the huge number of additional "degrees of freedom" that connectionist networks are afforded by virtue of the blanket claim of distributed representation across large numbers of uninterpreted units, as it is to any inherent virtues that connectionism has over symbolism in explaining cognitive phenomena. FOOTNOTES [1] There are many ways of studying implicit memory. The experiment I describe here is, I believe, a "classic" procedure, but by no means the only one. [2] A Psycoloquy reviewer of this paper suggested that it is not the individual units that are theoretical entities, but only the units as a general type. He explicitly compared the situation to that of statistical dynamics, in which the phenomena are said to result from the actions of large, but unspecified, numbers of molecules of a general type. The difference is, of course, that we have lots of independent evidence of the existence of molecules. We know quite a lot about their properties. The same cannot be said of units in connectionist networks. Their existence is posited SOLELY for the purpose of making the networks behave the way we want them to. There is no independent evidence of their properties or their existence at all. [3] Notice that a version of the Sorites paradox threatens here. There must come a point where the subtraction of a unit from the network would lead to a decrement in its performance, but typically connectionist researchers work well above this level in order to optimize learning speed and generalization. [4] One Psycoloquy referee suggested that units might correspond to small neural circuits rather than individual neurons. This might be so, but the evidential burden is clearly on the person who makes this proposal to find some convincing empirical evidence for it. [5] There may be a temptation to attempt to carry this through to the quantum level, and claim that it does not carry through at that level because of the physical impossibility of seeing subatomic particles. First of all, relying on our intuitions about the quantum world to illuminate other scientific spheres is a very dangerous move because it is there more than anywhere that our intuitions seem to fail. Despite this, the move would fail in any case because the impossibility at issue is merely PHYSICAL, not LOGICAL. In a world in which light turned out to be continuous rather than particulate, the argument would carry through perfectly well. Put less technically, we know WHERE to see subatomic particles, we just don't know HOW to see them. The same cannot be said for units in connectionist networks. They simply don't seem to refer to ANYTHING in the system being studied at all. REFERENCES Atkinson, R. C. & Shiffrin, R. M. (1971) The control of short-term memory. Scientific American 225:82-90. Baddeley (1992) Working memory. Science 255:556-559. Berwick, R. C. (1985) The acquisition of syntactic knowledge, MIT Press. Churchland, P. M. (1990) Cognitive activity in artificial neural networks. In: Thinking: An invitation to cognitive science (Vol. 3), ed. D. N. Osherson & E. E. Smith, MIT Press. Fodor, J. A. & Pylyshyn, Z. W. (1988). Connectionism and cognitive architecture: A critical analysis. Cognition 28:3-71. Hempel, C. G. (1952) Fundamentals of concept formation in empirical science. University of Chicago Press. Jacoby, L. L. (1991) A process dissociation framework: Separating automatic from intentional uses of memory. Journal of Memory & Language 30: 513-541. Lakatos, I. (1970). Falsification and the methodology of scientific research programmes. In: Criticism and the growth of knowledge, ed. I. Lakatos & A. Musgrave (Eds.), Cambridge University Press. Levitan, I. B. & Kaczmarek, L. K.. (1991). The neuron: Cell and molecular biology. New York: Oxford University Press. McClelland, J. L., Rumelhart, D. E., & Hinton, G. E. (1986) The appeal of parallel distributed processing. In: Parallel distributed processing: Explorations in the microstructure of cognition (vol. 1), ed. Rumelhart, D. E. & McClelland, J. L., MIT Press. Roediger, H. L. III (1990) Implicit memory: Retention without remembering. American Psychologist 45:1043-1056. Roediger, H. L., III, & McDermott, K. B. (1993) Implicit memory in normal human subjects. In: Handbook of neuropsychology (Vol. 8, pp. 63-131), ed. F. Boller & J. Grafman, Elsevier. Schacter, D. L. (1987) Implicit memory: History and current status. Journal of Experimental Psychology: Learning, Memory, and Cognition 13:501-518. Schacter, D. L. (1992). Understanding implicit memory: A cognitive neuroscience approach. American Psychologist 47:559- 569. Smolensky, P. (1988). On the proper treatment of connectionism. Behavioral and Brain Sciences 11:1-73. Tulving, E. (1985) How many memory systems are there? American Psychologist 40:385-398. From db10 at cus.cam.ac.uk Wed Apr 15 10:36:39 1998 From: db10 at cus.cam.ac.uk (David Brown) Date: Wed, 15 Apr 1998 15:36:39 +0100 Subject: POSTDOC - NEURAL MODELLER - CAMBRIDGE, UK Message-ID: <3.0.5.32.19980415153639.008f8dc0@pop.cus.cam.ac.uk> POSTDOCTORAL FELLOWSHIP - NEURAL MODELLER Laboratory of Computational Neuroscience, The Babraham Institute, Cambridge, UK. MODELLING A DUAL-FUNCTION NEURAL NETWORK This EU funded project linking laboratories in Cambridge, Edinburgh, Montpellier and Rome will involve the construction of mathematical models of oxytocin neurones, assessing model properties by computer simulation and analytical techniques where possible, and participation in the planning and execution of experimental tests of the models, in conjunction with neurobiologists (at Montpellier and Edinburgh), e.g. by devising critical experiments to discriminate between hypotheses. Random synaptic input is an important element of the environment of oxytocin neurones, which have two modes of activity corresponding to different physiological functions: firing in synchronised, high frequency bursts during lactation and parturition; and in a continuous firing pattern, responding to the imposed osmotic stress by a graded increase in mean activity. Models will probably be required at various levels of complexity, spanning the range from leaky integrator models to biophysical models with dendritic and somatic compartments. The postdoctoral research worker appointed will work closely with Jianfeng Feng and David Brown in the Lab. of Computational Neuroscience at Babraham, and in collaboration with mathematicians in the Physics Department, Rome 'La Sapienza' University (whose main focus will be assembling the single neurone models developed at Cambridge into networks) and neuroendocrinologists in Montpellier and Edinburgh. This is an exciting and novel project. The network's capacity for two distinct modes of action, both physiologically important, in response to different inputs will probably require the development of new models. Because of the low-dimensionality of system outputs (either tonic or pulsatile hormone release), sufficient good experimental data (e.g. simultaneous electrophysiological recordings, hormone release etc) can be collected within the project for thorough experimental calibration and testing of models. The biologists involved have between them an unrivalled experience and knowledge of the oxytocin system, and are pioneering cutting-edge experimental techniques for its study. The person appointed should have a PhD or equivalent research experience in biological, preferably neuronal or physiological modelling, with knowledge of analytical and simulation based techniques for assessing the behaviour of neuronal models, and relating the models to experimental data. A good first degree in a mathematically based subject and experience of biological computing will probably be required. The project will involve some travel between the four sites, so as to facilitate regular contact with the mathematicians and biologists involved. Salary in approx. range ?16,000-?26,000 per annum depending on qualifications and experience. The project will start in mid-1998, and continue in the first instance for 2 years. Further information from David Brown (+44 (0)1223 832312, Fax +44 (0)1223 837912, email db10 at cus.cam.ac.uk). Applications in the form of a CV and names and addresses of three referees to David Brown, Laboratory of Computational Neuroscience, Babraham Institute, Cambridge CB2 4AT as soon as possible and at the latest by 30th May 1998. From franz at homer.njit.edu Fri Apr 17 10:55:49 1998 From: franz at homer.njit.edu (Franz Kurfess) Date: Fri, 17 Apr 1998 10:55:49 -0400 Subject: CfP Special Issue "Neural Networks and Structured Knowledge" In-Reply-To: <40984038@toto.iv> Message-ID: <199804171455.KAA11297@vector.njit.edu> We received a number of requests to extend the deadline for the Special Issue "Neural Networks and Structured Knowledge" in "Applied Intelligence", and decided to revise the schedule, in effect pushing all dates back by two months. Here is the new schedule: Revised Schedule Paper submission deadline: July 1, 1998 Review decision by: August 31, 1998 Final manuscript due: September 31, 1998 Tentative publication date: January 1999 I'm appending the Call for Contributions for your reference. Best regards, Franz Kurfess, Guest Editor Special Issue "Neural Networks and Structured Knowledge" in Applied Intelligence: The International Journal of Artificial Intelligence, Neural Networks, and Complex Problem-Solving Techniques Call for Contributions The submission of papers is invited for a special issue on "Neural Networks and Structured Knowledge" of the Applied Intelligence Journal. Issue Theme The representation and processing of knowledge in computers traditionally has been concentrated on symbol-oriented approaches, where knowledge items are associated with symbols. These symbols are grouped into structures, reflecting the important relationships between the knowledge items. Processing of knowledge then consists of manipulation of the symbolic structures, and the result of the manipulations can be interpreted by the user. Whereas this approach has seen some remarkable successes, there are also domains and problems where it does not seem adequate. Some of the problems are computational complexity, rigidity of the representation, the difficulty of reconciling the artificial model with the real world, the integration of learning into the model, and the treatment of incomplete or uncertain knowledge. Neural networks, on the other hand, have advantages that make them good candidates for overcoming some of the above problems. Whereas approaches to use neural networks for the representation and processing of structured knowledge have been around for quite some time, especially in the area of connectionism, they frequently suffer from problems with expressiveness, knowledge acquisition, adaptivity and learning, or human interpretation. In the last years much progress has been made in the theoretical understanding and the construction of neural systems capable of representing and processing structured knowledge in an adequate way, while maintaining essential capabilities of neural networks such as learning, tolerance of noise, treatment of inconsistencies, and parallel operation. The theme of this special issue comprises * the investigation of the underlying theorecical foundations, * the implementation and evaluation of methods for representation and processing of structured knowledge with neural networks, and * applications of such approaches in various domains. Topics of Interest The list below gives some examples of intended topics. * Concepts and Methods: o extraction, injection and refinement of structured knowledge from, into and by neural networks o inductive discovery/formation of structured knowledge o combining symbolic machine learning techniques with neural lerning paradigms to improve performance o classification, recognition, prediction, matching and manipulation of structured information o neural methods that use or discover structural similarities o neural models to infer hierachical categories o structuring of network architectures: methods for introducing coarse-grained structure into networks, unsupervised learning of internal modularity * Application Areas: o medical and technical diagnosis: discovery and manipulation of structured dependencies, constraints, explanations o molecular biology and chemistry: prediction of molecular structure unfolding, classification of chemical structures, DNA analysis o automated reasoning: robust matching, manipulation of logical terms, proof plans, search space reduction o software engineering: quality testing, modularisation of software o geometrical and spatial reasoning: robotics, structured representation of objects in space, figure animation, layouting of objects o other applications that use, generate or manipulate structures with neural methods: strucures in music composition, legal reasoning, architectures, technical configuration, ... The central theme of this issue will be the treatment of structured information using neural networks, independent of the particular network type or processing paradigm. Thus the theme is orthogonal to the question of connectionist/symbolic integration, and is not intended as a continuation of the more philosphically oriented discussion of symbolic vs. subsymbolic representation and processing. Submission Process Prospective authors should send an electronic mail message indicating their intent to submit a paper to the guest editor of the special issue, Franz J. Kurfess (kurfess at cis.njit.edu). This message should contain a preliminary abstract and three to five keywords. Six hard copies of the final manuscript should be sent to the guest editor (not to the Applied Intelligence Editorial office): Prof. Franz J. Kurfess New Jersey Institute of Technology Phone: (973) 596 5767 Department of Computer and Information Science Fax: (973) 596 5777 University Heights Email: kurfess at cis.njit.edu Newark, NJ 07102-1982 WWW: http://www.cis.njit.edu/~franz To speed up the reviewing process, authors should also send a PostScript version of the paper via email to the guest editor. Prospective authors can find further information about the journal on the home page http://kapis.www.wkap.nl/journalhome.htm/0924-669X Schedule Paper submission deadline: July 1, 1998 Review decision by: August 31, 1998 Final manuscript due: September 31, 1998 Tentative publication date: January 1999 From mieko at hip.atr.co.jp Fri Apr 17 07:11:26 1998 From: mieko at hip.atr.co.jp (Mieko Namba) Date: Fri, 17 Apr 1998 20:11:26 +0900 Subject: CALL FOR PAPERS [Neural Networks 1999 Special Issue] Message-ID: <199804171111.UAA04869@mailhost.hip.atr.co.jp> Dear members, We are glad to inform you that the Japanese Neural Networks Society will edit the NEURAL NETWORKS 1999 Special Issue as below. NEURAL NETWORKS is an official international compilation of the Journal of the International Neural Networks Society, the European Neural Networks Society and the Japanese Neural Networks Society. We are looking forward to receiving your contributions. Mitsuo Kawato Co-Editor-in-Chief Neural Networks (ATR Human Information Proc. Res. Labs.) ****************************************************************** CALL FOR PAPERS ****************************************************************** Neural Networks 1999 Special Issue "Organisation of Computation in Brain-like Systems" ****************************************************************** Co-Editors: Professor Gen Matsumoto, BSI, RIKEN, Japan Professor Edgar Koerner, HONDA R&D, Europe Dr. Mitsuo Kawato, ATR Human Information Processing Res. Labs., Japan Submission: Deadline for submission: December 1st, 1998 Notification of acceptance: March 1st, 1999 Format: as for normal papers in the journal (APA format) and no longer than 10,000 words Address for Papers: Dr. Mitsuo Kawato ATR Human Information Processing Research Laboratories 2-2 Hikaridai, Seika-cho Soraku-gun, Kyoto 619-0288, Japan. ****************************************************************** In the recent years, neuroscience has made a big leap forward regarding both investigation methodology and insights in local mechanisms of processing of sensory information in the brain. The fact that we still do not know much better than before what happens in the brain when one recognises a familiar person, or moves around navigating seemingly effortless through a busy street, points to the fact that our models still do not describe essential aspects of how the brain organises computation. The investigation of the behaviour of fairly homogeneous ANS (artificial neural systems) composed of simple elementary nodes fostered the awareness that architecture matters: Algorithms implemented by the respective neural system are expressed by its architecture. Consequently, the focus is shifting to better understanding of the architecture of the brain and of its subsystems, since the structure of those highly modularised systems represents the way the brain organises computation. Approaching the algorithms expressed by those architectures may offer us the capability to not only understand the representation of knowledge in a neural system made under well defined constraints, but to understand the control that forces the neural system to make representations of behaviourally relevant knowledge by generating dynamic constraints. This special issue will bring together invited papers and contributed articles that illustrate the shifting emphasis in neural systems modelling to more neuroarchitecture-motivated systems that include this type of control architectures. Local and global control algorithms for organisation of computation in brain-like systems cover a wide field of topics. Abduction of control principles inherent in the architectures that mediate interaction within the cortex, between cortex -thalamus, cortex-hippocampus and other parts of the limbic system is one of the targets. Of particular importance are the rapid access to stored knowledge and the management of conflicts in response to sensory input, the coding and representation in a basically asynchronous mode of processing, the decomposition of problems into a reasonable number of simpler sub-problems, and the control of learning -- including the control which specifies what should be learned, and how to integrate the new knowledge into the relational architecture of the already acquired knowledge representation. Another target of that approach is the attempt to understand how these controls and the respective architectures emerged in the process of self-organisation in the phylogenetic and ontogenetic development. Setting the cognitive behaviour of neural systems in the focus of investigation is a prerequisite for the described approach that will promote both creating computational hypotheses for neurobiology and implementing robust and flexible computation in ANS. ****************************************************************** end. ========================================================= Mieko Namba Secretary to Dr. Mitsuo Kawato Editorial Administrator of NEURAL NETWORKS ATR Human Information Processing Research Laboratories 2-2 Hikaridai, Seika-cho, Soraku-gun, Kyoto 619-0288, Japan TEL +81-774-95-1058 FAX +81-774-95-1008 E-MAIL mieko at hip.atr.co.jp ========================================================= From pmitra at bell-labs.com Mon Apr 20 22:58:09 1998 From: pmitra at bell-labs.com (Partha Mitra) Date: Mon, 20 Apr 1998 22:58:09 -0400 Subject: Please distribute Message-ID: <353C0B41.5C03@bell-labs.com> _______________________ Analysis of Neural Data _______________________ Modern methods and open issues in the analysis and interpretation of multi-variate time-series and imaging data in the neurosciences ___________________________________________________ >> 16 August - 29 August 1998 >> Marine Biological Laboratories - Woods Hole, MA ___________________________________________________ A working group of scientists committed to quantitative approaches to problems in neuroscience will focus their efforts on experimental and theoretical issues related to the analysis of large, single- and multi-channel data sets. The motivation for the work group is based on issues that arise in two complimentary areas critical to an understanding of brain function. The first involves advanced signal processing methods, particularly those appropriate for emerging multi-site recording techniques and noninvasive imaging techniques. The second involves the development of a calculus to study the dynamical behavior of nervous systems and the computations they perform. A distinguishing feature of the work group will be the close collaboration between experimentalists and theorists, particularly with regard to the analysis of data and the planning of experiments. The work group will have a limited number of research lectures, supplemented by tutorials on relevant computational, experimental, and mathematical techniques. This work group is a means to critically evaluate techniques for the processing of multi-channel data, of which imaging forms an important category. Such techniques are of fundamental importance for basic research and medical diagnostics. We will establish a repository of these techniques, along with benchmarks, to insure the rapidly dissemination of modern analytical techniques throughout the neuroscience community. The work group will convene on a yearly basis. For 1997, we propose to focus on topics that fall under the rubric of multivariate time-series. * Analysis of point processes, e.g., spike trains. Measures of correlation and variability, and their interpretation. * Analysis of continuous processes, e.g., field potential, optical imaging, fMRI, and MEG, and the recording of behavioral output, e.g., vocalizations. * Problems that involve both point and continuous processes, e.g., the linear and nonlinear functional relations between spike trains and sensory input and motor output. Participants: Twenty five participants, both experimentalists and theorists. Experimentalists are specifically encouraged to bring data records to the work group; appropriate computational facilities will be provided. The work group will further take advantage of interested investigators and course faculty concurrently present at the MBL. We encourage graduate students and postdoctoral fellows as well as senior researchers to apply. Participant Fee: $200. Support: National Institutes of Health - NIMH, NIA, NIAAA, NICHD/NCRR, NIDCD, NIDA, and NINDS. Organizers: David Kleinfeld (UCSD) and Partha P. Mitra (Caltech and Bell Laboratories). Website: http://www-physics.ucsd.edu/research/neurodata Application: Send a copy of your curriculum vita, together with a cover letter that contains a brief (ca. 200 word) paragraph on why you wish to attend the work group and a justified request for any financial aid, to: Ms. Jean B. Ainge Bell Laboratories, Lucent Technologies 700 Mountain Avenue 1D-427 Murray Hill, NJ 07974 908-582-4702 (fax) or The MBL is an EEO AAI. Graduate students and postdoctoral fellows are encouraged to include a brief letter of support from their research advisor. Financial assistance: Assistance for travel, accommodations, and board is available based on need. Applications must be received by 18 May 1998. Participants will be notified by 25 May Links to Archives for Neurosciences can be found at: http://www-physics.ucsd.edu/research/neurodata/NSarchive2.html From Wulfram.Gerstner at epfl.ch Wed Apr 22 09:49:39 1998 From: Wulfram.Gerstner at epfl.ch (Wulfram Gerstner) Date: Wed, 22 Apr 1998 15:49:39 +0200 (MET DST) Subject: preprints_on_spiking_neurons Message-ID: <199804221349.PAA06196@mantrasun8.epfl.ch> Three review papers on Neural Networks with Spiking Neurons can be retrieved from the following web page. http://diwww.epfl.ch/lami/team/gerstner/wg_pub.html --------------------------------------------------- I. SPIKING NEURONS. (W. Gerstner) This tutorial paper gives a review of several models of spiking neurons (integrate-and-fire; Hodgkin-Huxley; spike response model). (55 pages) II. POPULATIONS OF SPIKING NEURONS. (W. Gerstner) The second paper develops the mathematical framework to describe populations of spiking neurons. Specific topics are (i) the rapid response of populations of spiking neurons to changes in the input. (ii) exact stability conditions for perfectly synchronized locked solutions as well as (iii) the stability of incoherent firing activity in the presence of noise and transmission delays. (38 pages) III. HEBBIAN LEARNING OF PULSE TIMING IN THE BARN OWL AUDITORY SYSTEM. (W. Gerstner, R. Kempter, J.L. van Hemmen and H. Wagner) A correlation based learning rule based on spike timing is discussed and applied to the problem of coincidence detection and sound localization in the barn owl auditory system. Specific topics are (i) the relation of spike-based and rate-based learning (ii) delay line selection by learning (iii) phase locking in the auditory system. (26 pages) ----------------------------------------------------- The three papers are preprints of book chapters. The book entitled Pulsed Neural Nets edited by W. Maas and C. Bishop will appear in October/November 98 (MIT Press). From dror at coglit.soton.ac.uk Wed Apr 22 18:28:52 1998 From: dror at coglit.soton.ac.uk (Itiel Dror) Date: Wed, 22 Apr 1998 23:28:52 +0100 (BST) Subject: Position at Southampton University Message-ID: I would appreciate it very much if you could please post the following job announcement. Thank you. Itiel #======================================================================# | Itiel E. Dror, Ph.D. http://www.cogsci.soton.ac.uk/~dror/ | | Department of Psychology dror at coglab.psy.soton.ac.uk | | University of Southampton Office 44 (0)1703 594519 | | Highfield, Southampton Lab. 44 (0)1703 594518 | | England SO17 1BJ Fax. 44 (0)1703 594597 | #======================================================================# ******************************************************************************* UNIVERSITY OF SOUTHAMPTON DEPARTMENT OF PSYCHOLOGY FULL PROFESSOR IN COGNITIVE PSYCHOLOGY Applications are invited for a Full Professor in Cognitive Psychology in the Department of Psychology tenable as soon as possible. In a recent review of our research activity the University identified the further strengthening of the Cognitive Psychology base within the Department as a strategic priority. As a result the University has provided additional resources to support developments in this field. The established Professorship which falls vacant in September this year has been specifically designated as a post in Cognitive Psychology. In addition, two new junior position have been created to support the appointment of the new Full Professor. The junior positions will be advertised only after the appointment of the Full Professor in order to promote complementarity amongst our Cognitive Psychology grouping. Substantial funds will be set aside for the purchase of equipment to support the research of the successful applicant. The Department has recently moved into refurbished premises in the Shackleton Building which provide flexible research space with extensive laboratory facilities for experimental work. The new Professor in Cognitive Psychology will be given lighter teaching and administrative duties during their first year than would normally be the case. These new posts in Cognitive Psychology are part of a larger package of new HEFCE funded appointments. There will also be a new Full Professor and a new junior position in Social Psychology, two senior positions (Health and Developmental Psychology) and a junior position in Human Learning and Behaviour Analysis. Each post has been targeted to play a specific role in relation to one of the new Research Groups identified during the review mentioned above. Technical and clerical support within the Department is also being overhauled. These changes will lead to a further period of growth and development within the Department and further strengthen what is already a lively and effective research team. At the same time they will increase the scope of teaching at both the undergraduate and postgraduate level. These developments also provide a clear indication of the University of Southampton's support for Psychology's commitment to become a major center of excellence in research and teaching. The new posts in Cognitive Psychology will strengthen what is already a productive group of young cognitive scientists working within the Department. The successful candidate will be expected to provide leadership in this area and so play a major role in the development of the recently established Cognitive Psychology Research Group. An essential requirement is that he or she will have the capacity to facilitate the research of others. Collaboration with colleagues within and outside the Cognitive Psychology grouping will be encouraged. The appointee's research could be in any area of Cognitive Psychology. However, he or she will have an outstanding track-record of achievement in theoretically driven empirical research of a fundamental nature evidenced by publication in the leading international refereed journals. The Department would welcome informal enquiries and visits. Potential applicants should contact the Head of Department, Professor Edmund Sonuga-Barke, on (01703) 594606 or esb at psy.soton.ac.uk. or to Dr Itiel Dror (dror at coglab.psy.soton.ac.uk) or Dr Sarah Stevenage (svs1 at soton.ac.uk) From dld at cs.monash.edu.au Thu Apr 23 07:28:11 1998 From: dld at cs.monash.edu.au (David L Dowe) Date: Thu, 23 Apr 1998 21:28:11 +1000 Subject: CFPs: Information theory in biology, Jan 99, Hawaii Message-ID: <199804231128.VAA11139@dec11.cs.monash.edu.au> Dear All, Apologies for cross-postings. In short, if you're interested in MML or MDL or Akaike's Information Criterion or information theory and you're interested in biology, and you'd like to go to Hawaii in January 1999, and you can get a paper ready by July 1998, then store this mail away and bookmark the site http://www.cs.monash.edu.au/~dld/PSB99/PSB99.Info.CFPs.html and read on. =================================================================== This is the Call For Papers for the 4th Pacific Symposium on BioComputing (PSB99, 1999) conference track on "Information-theoretic approaches to biology". PSB-99 will be held from 4-9 January, 1999, in Mauni Lani on the Big Island of Hawaii. Track Organisers: David L. Dowe (dld at cs.monash.edu.au) and Klaus Prank. WWW site: http://www.cs.monash.edu.au/~dld/PSB99/PSB99.Info.CFPs.html . Specific technical area to be covered by this track: Approaches to biological problems using notions of information or complexity, including methods such as Algorithmic Probability, Minimum Message Length and Minimum Description Length. Two possible applications are (e.g.) protein folding and biological information processing. Kolmogorov (1965) and Chaitin (1966) studied the notions of complexity and randomness, with Solomonoff (1964), Wallace (1968) and Rissanen (1978) applying these to problems of statistical and inferential learning (and ``data mining'') and to prediction. The methods of Solomonoff, Wallace and Rissanen have respectively come to be known as Algorithmic Probability (ALP), Minimum Message Length (MML) and Minimum Description Length (MDL). All of these methods relate to information theory, and can also be thought of in terms of Shannon's information theory, and can also be thought of in terms of Boltzmann's thermo-dynamic entropy. An MDL/MML perspective has been suggested by a number of authors in the context of approximating unknown functions with some parametric approximation scheme (such as a neural network). The designated measure to optimize under this scheme combines an estimate of the cost of misfit with an estimate of the cost of describing the parametric approximation (Akaike 1973, Rissanen 1978, Barron and Barron 1988, Wallace and Boulton, 1968). This track invites all original papers of a biological nature which use notions of information and/or information-theoretic complexity, with no strong preference as to what specific nature. Such work has been done in problems of, e.g., protein folding and DNA string alignment. As we shortly describe in some detail, such work has also been done in the analysis of temporal dynamics in biology such as neural spike trains and endocrine (hormonal) time series analysis using the MDL principle in the context of neural networks and context-free grammar complexity. To elaborate on one of the relevant topics above, in the last three years or so, there has been a major focus on the aspect of timing in biological information processing ranging from fields such as neuroscience to endocrinology. The latest work on information processing at the single-cell level using computational as well as experimental approaches reveals previously unimagined complexity and dynamism. Timing in biological information processing on the single-cell level as well as on the systems level has been studied by signal-processing and information-theoretic approaches in particular in the field of neuroscience (see for an overview: Rieke et al. 1996). Using such approaches to the understanding of temporal complexity in biological information transfer, the maximum information rates and the precision of spike timing to the understanding of temporal complexity in biological information transfer, the maximum information rates and the precision of spike timing could be revealed by computational methods (Mainen and Sejnowski, 1995; Gabbiani and Koch 1996; Gabbiani et al., 1996). The examples given above are examples of some possible biological application domains. We invite and solicit papers in all areas of (computational) biology which make use of ALP, MDL, MML and/or other notions of information and information-theoretic complexity. In problems of prediction, as well as using "yes"/"no" predictions, we would encourage the authors to consider also using probabilistic prediction, where the score assigned to a probabilistic prediction is given according to the negative logarithm of the stated probability of the event. Further comments re PSB-99 : ---------------------------- PSB99 will publish accepted full papers in an archival Proceedings. All contributed papers will be rigorously peer-reviewed by at least three referees. Each accepted full paper will be allocated up to 12 pages in the conference Proceedings. The best papers will be selected for a 30-minute oral presentation to the full assembled conference. Accepted poster abstracts will be distributed at the conference separately from the archival Proceedings. To be eligible for proceedings publication, each full paper must be accompanied by a cover letter stating that it contains original unpublished results not currently under consideration elsewhere. See http://www.cgl.ucsf.edu/psb/cfp.html for more information. IMPORTANT DATES: Full paper submissions due: July 13, 1998 Poster abstracts due: August 22, 1998 Notification of paper acceptance: September 22, 1998 Camera-ready copy due: October 1, 1998 Conference: January 4 - 9, 1999 More information about the "Information-theoretic approaches to biology" track, including a sample list of relevant papers is available on the WWW at http://www.cs.monash.edu.au/~dld/PSB99/PSB99.Info.CFPs.html . More information about PSB99 is available from http://www.cgl.ucsf.edu/psb/cfp.html For further information, e-mail Dr. David Dowe, dld at cs.monash.edu.au or e-mail Dr. Klaus Prank, ndxdpran at rrzn-serv.de . This page was put together by Dr. David Dowe, School of Computer Science and Softw. Eng., Monash University, Clayton, Vic. 3168, Australia e-mail: dld at cs.monash.edu.au Fax: +61 3 9905-5146 http://www.csse.monash.edu.au/~dld/ and Dr. Klaus Prank, Abteilung Klinische Endokrinologie Medizinische Hochschule Hannover Carl-Neuberg-Str. 1 D-30623 Hannover Germany e-mail: ndxdpran at rrzn-serv.de Tel.: +49 (511) 532-3827 Fax.: +49 (511) 532-3825 http://sun1.rrzn-user.uni-hannover.de/~ndxdpran/ From Paul.Keller at pnl.gov Thu Apr 23 13:26:12 1998 From: Paul.Keller at pnl.gov (Keller, Paul E) Date: Thu, 23 Apr 1998 10:26:12 -0700 Subject: Job Announcement: Cognitive Systems Engineer at Battelle Message-ID: <7A8CF1DC6A9DD0118EA400A024BF29DA0229BB89@pnlmse2.pnl.gov> Cognitive Systems Engineer Battelle, a leading provider of technology solutions, has immediate need for a research engineer to join their cognitive systems initiative in their Columbus, Ohio, USA facility. The new position will provide technical support to a multi-year corporate project applying adaptive/cognitive information technology to applications in emerging technology areas. The position requires a B.S./M.S. in Computer and Information Science, Electrical Engineering, or related field with a specialization or experience in artificial neural networks, fuzzy logic, evolutionary computing/genetic algorithms, and statistical methods. In addition, the position requires intimate knowledge of Matlab, C/C++ language and object-oriented programming. Oral, written, and interpersonal communications skills are essential to this highly interactive position. The applicant selected will be subject to a security investigation and must meet eligibility requirements for access to classified information. Battelle offers competitive salaries, comprehensive benefits, and opportunities for professional development. Qualified candidates are invited to send their resumes to Battelle, Dept. J-92, 505 King Avenue, Columbus, OH 43201-2693 or e-mail them to priddy at battelle.org. Battelle is an Equal Opportunity/Affirmative Action Employer M/F/D/V. To find out more information about Battelle, try http://www.battelle.org. *** PLEASE RESPOND TO THE ADDRESS GIVEN ABOVE OR E-MAIL KEVIN PRIDDY AT priddy at battelle.org *** From n at predict.com Fri Apr 24 12:10:53 1998 From: n at predict.com (Norman Packard) Date: Fri, 24 Apr 1998 10:10:53 -0600 Subject: Job Announcement: research/software at Prediction Company Message-ID: <199804241610.KAA07707@seldon> PREDICTION COMPANY RESEARCH/SOFTWARE POSITION IN NONLINEAR MODELING OF FINANCIAL MARKETS April, 1998 Prediction Company is a small firm based in Santa Fe, NM, utilizing nonlinear forecasting technologies for prediction and computerized trading of financial instruments. We are seeking someone that can play a strong role in both research and research software. The basic task is to build models based on historical data to trade in financial markets. Responsibilities include application of existing technology, research and development of new technology, and participation in the design and building of an advanced software platform for prediction, trading, and risk control. The successful applicant for this job will have a Ph.D. in statistics, econonimcs, computer science, physics, mathematics, or a related field. Experience using time series modeling, machine learning, and statistical and numerical anlysis is highly valuable. Software experience is essential, particularly desirable in C++, S+ or related languages. Familiarity with finance is highly desirable. Experience with real data is also highly desirable -- the nastier the better. We are willing to consider a range of experience levels, including recent Ph.D.'s. The applicant should be willing to work in close collaboration with other researchers and software developers, and should be willing to take on what is unquestionably the most challenging but lucrative forecasting problem in existence. Prediction Company offers a relaxed and informal work environment. We are located in an historic three story building in the Guadalupe commercial district. Our offices include a full kitchen and roof deck. We are within easy walking distance of many cafes, restaurants and the historical central plaza of Santa Fe. For further information check out our web page at www.predict.com. Applicants should email resumes to Laura Barela at laura at predict.com (postscript or ascii) or send by US mail to: Prediction Company Attn: Recruiting 236 Montezuma Avenue Santa Fe, NM 87501 From niall at zeus.csis.ul.ie Mon Apr 27 06:28:39 1998 From: niall at zeus.csis.ul.ie (Niall Griffith) Date: Mon, 27 Apr 1998 11:28:39 +0100 Subject: IEE Colloqiuim - Neural Nets and MultiMedia Message-ID: <9804271028.AA17535@zeus.csis.ul.ie> I am sorry if you receive this twice - according to a message that I have received it has not been delivered to you so I am trying again Niall Griffith My message was..... -------------------------------------------------------------- Please pass this on to anyone or any group you think may be interested. ============================================================== IEE Colloquium on "Neural Networks in Multimedia Interactive Systems" Thursday 22 October 1998, Savoy Place, London. Call for Papers - --------------- The IEE are holding a colloquium at Savoy Place on the use of neural network models in multimedia systems. This is a developing field of importance to both Multimedia applications developers who want to develop more responsive and adaptive systems as well as to neural network researchers. The aim of the colloquium is to present a range of current neural network applications in the area of interactive multimedia. The aim is cover a range of topics including learning, intelligent agents within multimedia systems, data mining, image processing and intelligent application interfaces. Invited Speakers: - ----------------- Bruce Blumberg, MIT Media Lab. Jim Austin, York. Russell Beale, Birmingham. Call For Papers - --------------- Submissions are invited in any (but not exclusively) of the following areas: Adaptive and plastic behaviour in multi-media systems Concept and behaviour learning and acquisition Browsing mechanisms Preference and strategy identification and learning Data mining Image processing in multimedia systems Cross modal and media representations and processes Intelligent agents Interested parties are invited to submit a two page (maximum) abstract of their proposed talk to either Dr. Niall Griffith, Department of Computer Science and Information Science, University of Limerick, Limerick, Ireland. email: niall.griffith at ul.ie Telephone: +353 61 202785 Fax: +353 61 330876 or Professor Nigel M Allinson Dept. of Elec. Eng. & Electronics UMIST PO Box 88 Manchester, M60 1QD, UK Voice: (+44) (0) 161-200-4641 Fax: (+44) (0) 161-200-4781/4 Internet: allinson at umist.ac.uk Timetable: - ---------- 29th April: Deadline for talk submissions 15th June: Authors notified. 24th November: Colloquium at IEE, Savoy Place, London ===================================================== - --RAB11527.891386128/oz.memphis.edu-- ------- End of forwarded message ------- From harnad at coglit.soton.ac.uk Mon Apr 27 09:24:08 1998 From: harnad at coglit.soton.ac.uk (Stevan Harnad) Date: Mon, 27 Apr 1998 14:24:08 +0100 (BST) Subject: Contribute to Ongoing Psyc Commentary on Green Message-ID: There is lively Commentary on Green's target article appearing in Psycoloquy, a refereed electronic journal sponsored by the American psychological Association. Further Commentary is invited. (All submissions are refereed.) URLs: US: http://www.princeton.edu/~harnad/psyc.html UK: http://www.cogsci.soton.ac.uk/psyc Address for submitting commentaries: psyc at pucc.princeton.edu Instructions at bottom of this message, preceded by latest commentary. Green, CD. Are Connectionist Models Theories of Cognition? PSYCOLOQUY 9(04) Tuesday 14 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.04.connectionist-explanation.1.green Orbach, J. Do Wires Model Neurons? PSYCOLOQUY 9(05) Wednesday 15 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.05.connectionist-explanation.2.orbach O'Brien, GJ. The Role of Implementation in Connectionist Explanation. PSYCOLOQUY 9(06) Sunday 19 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.06.connectionist-explanation.3.obrien Green, CD. Lashley's Lesson Is Not Germane. Reply to Orbach PSYCOLOQUY 9(07) Wednesday 22 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.07.connectionist-explanation.4.green Green, CD. Problems with the Implementation Argument. Reply to O'Brien PSYCOLOQUY 9(08) Saturday 25 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.08.connectionist-explanation.5.green Young, ME. Are Hypothetical Constructs Preferred Over Intervening Variables? PSYCOLOQUY 9(09) Monday 27 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.09.connectionist-explanation.6.young Grainger, J. & Jacobs, AM. Localist Connectionism Fits the Bill PSYCOLOQUY 9(09) Monday 27 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.10.connectionist-explanation.7.grainger ---------- psycoloquy.98.9.10.connectionist-explanation.7.grainger Mon 27 Apr 1998 ISSN 1055-0143 (6 paragraphs, 8 references, 153 lines) PSYCOLOQUY is sponsored by the American Psychological Association (APA) Copyright 1998 Jonathan Grainger LOCALIST CONNECTIONISM FITS THE BILL Commentary on Green on Connectionist-Explanation Jonathan Grainger Centre de Recherche en Psychologie Cognitive, CNRS Universite de Provence Aix-en-Provence France grainger at newsup.univ-mrs.fr Arthur M. Jacobs Dept. of Psychology Philips University of Marburg, Marbug, Germany jacobsa at mailer.uni-marburg.de ABSTRACT: Green (1998) restates a now standard critique of connectionist models: they have poor explanatory value as a result of their opaque functioning. However, this problem only arises in connectionist models that use distributed hidden unit representations, and is NOT a feature of localist connectionism. Indeed, Green's critique reads as an appeal for the development of localist connectionist models as an excellent starting point for building a unified theory of human cognition. 1. First, if we agree that theory development in psychological science is ready for the shift from prequantitative verbal-boxological modeling toward more formal modeling efforts, then the kinds of questions we should be asking are: What kind of quantitative modeling is appropriate? How should we evaluate its appropriateness? In other words, the verbal theories of human memory discussed by Green (1998) are not a serious alternative to whatever connectionism might offer. They are at best a starting point for developing more formal accounts of human memory. We have recently argued that localist connectionism provides a promising framework for such an endeavor (Grainger & Jacobs, 1998). 2. Green (1998), as well as many other critics of connectionism, appears to use the term connectionism as synonymous with trainable networks with hidden units (often called PDP models, and typically trained with backpropagation, Rumelhart, Hinton, & Williams, 1986). Many connectionist models do not include hidden units. Some of these are trainable (with Hebbian learning, for example), and some are hardwired (e.g., McClelland & Rumelhart's, 1981, interactive activation model). We refer to any connectionist model in which all processing units can be unambiguously assigned a meaningful interpretation as "localist connectionist." Note that, as in all connectionist models, all processing units in localist connectionist models are identical; it is only their position in the network that guarantees their unique interpretation. The modeler can artificially label each of these units in order to facilitate interpretation of network activity. 3. Grainger and Jacobs (1998) analyzed the advantages of adopting a localist connectionist approach as opposed to the currently more popular PDP approach. Here we will discuss only those points relevant to the issues raised by Green (1998). Green identifies the close connection between theoretical and observable entities as a critical feature of traditional scientific theories. One must be able to link transparently the theoretical entities of the theory to the observable entities in the target world in order to achieve explanatory adequacy. Without examining the extent to which this is fails to be a feature of PDP models, it should be clear from the above discussion that localist connectionist models do provide this transparent link. Units in localist connectionist models do refer to relatively uncontroversial aspects of the target world. They represent the categories (such as letters and words) that the brain has learned from repeated exposure to the environment. 4. As noted by Jacobs, Rey, Ziegler, and Grainger (1998), transparency will always tend to diminish as models become more complex. Jacobs et al. conclude, however, that algorithmic models of the localist connectionist variety may offer the best trade-off between clarity/transparency and formality/precision. It is the increased level of precision that allows localist connectionist models to achieve greater descriptive adequacy (Jacobs & Grainger, 1994) without sacrificing explanatory adequacy. 5. Apart from greater explanatory and descriptive adequacy, localist connectionist models offer a simple means of quantifying pre-existing verbal-boxological models that have already stood the test of extensive empirical research. Referring to this point, Page and Norris (1998) speak of a symbiosis between verbal theorizing and quantitative modeling. Furthermore, the principle of nested modeling has been readily applied with localist connectionist models. Adopting this approach facilitates the process of model-to-model comparison. Models differing by a single feature (e.g., interactivity, Jacobs & Grainger, 1992), can be compared, and different variants of the model can compete in strong inference studies (e.g., Dijkstra & van Heuven, 1998). 6. Finally, localist connectionist models, using the same simple processing units and activation functions, provide a unified explanation for phenomena observed in the different subdomains of human cognition. The general principles that govern processing in all localist models (e.g., similarity based parallel activation, lateral inhibition) can also be isolated and analyzed in an easily interpretable manner (see e.g., Grainger & Jacobs, in press). We therefore conclude that localist connectionism provides an excellent starting point for the development of a unified theory of human cognition. REFERENCES Dijkstra, T. & van Heuven, W.J.B. (1998). The BIA model and bilingual word recognition. In J. Grainger & A.M. Jacobs (Eds.), Localist connectionist approaches to human cognition. Mahwah, NJ.: Erlbaum. Grainger, J. & Jacobs, A.M. (1998). On localist connectionism and psychological science. In J. Grainger & A.M. Jacobs (Eds.), Localist connectionist approaches to human cognition. Mahwah, NJ.: Erlbaum. Grainger, J. & Jacobs, A.M. (1998). Temporal integration of information in orthographic priming. Visual Cognition, in press. Green, CD. (1998) Are Connectionist Models Theories of Cognition? PSYCOLOQUY 9(4) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.04.connectionist-explanation.1.green Jacobs, A.M. & Grainger, J. (1992). Testing a semistochastic variant of the interactive activation model in different word recognition experiments. Journal of Experimental Psychology: Human Perception and Performance, 18, 1174-1188. Jacobs, A. M., & Grainger, J. (1994). Models of visual word recognition: Sampling the state of the art. Journal of Experimental Psychology: Human Perception and Performance, 20, 1311-1334. Jacobs, A.M., Rey, A., Ziegler, J.C, & Grainger, J. (1998). MROM-P: An interactive activation, multiple read-out model of orthographic and phonological processes in visual word recognition. In J. Grainger & A.M. Jacobs (Eds.), Localist connectionist approaches to human cognition. Mahwah, NJ.: Erlbaum. McClelland, J. L. & Rumelhart, D. E. (1981). An interactive activation model of context effects in letter perception: Part I. An account of basic findings. Psychological Review, 88, 375-407. Page, M. & Norris, D. (1998). Modeling immediate serial recall with a localist implementation of the primacy model. In J. Grainger & A.M. Jacobs (Eds.), Localist connectionist approaches to human cognition. Mahwah, NJ.: Erlbaum. Rumelhart, D.E., Hinton, G.E. & Williams, R.J. (1986). Learning internal represenatations by error propagation. In D.E. Rumelhart, J.L. McClelland, & the PDP research group, Parallel distributed processing: Explorations in the microstructure of cognition (Vol. 1). Cambridge, MA: Bradford Books. INSTRUCTIONS FOR PSYCOLOQUY 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 50,000. 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. Accepted PSYCOLOQUY target articles have been judged by 5-8 referees to be appropriate for Open Peer Commentary, the special service provided by PSYCOLOQUY to investigators in psychology, neuroscience, behavioral biology, cognitive sciences and philosophy who wish to solicit multiple responses from an international group of fellow specialists within and across these disciplines to a particularly significant and controversial piece of work. If you feel that you can contribute substantive criticism, interpretation, elaboration or pertinent complementary or supplementary material on a PSYCOLOQUY target article, you are invited to submit a formal electronic commentary. 1. Before preparing your commentary, please examine recent numbers of PSYCOLOQUY if not familiar with the journal. 2. Commentaries should preferably be up to ~200 lines (~1800 words) 3. Please provide a title for your commentary. 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 commentary should also have a brief (~100 word) abstract 4. All paragraphs should be numbered consecutively. Line length should not exceed 72 characters. The commentary 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, preparation and editing. Target article length should preferably be up to 1200 lines [c. 10,000 words]. All target articles, commentaries and responses must have (1) a short abstract (up to 200 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 URLs for retrieving full texts of target articles: http://cogsci.soton.ac.uk/psyc http://www.princeton.edu/~harnad/psyc.html gopher://gopher.princeton.edu:70/11/.libraries/.pujournals ftp://ftp.princeton.edu/pub/harnad/Psycoloquy ftp://cogsci.soton.ac.uk/pub/harnad/Psycoloquy news:sci.psychology.journals.psycoloquy Anonymous ftp archive is DIRECTORY pub/harnad/Psycoloquy HOST ftp.princeton.edu From aapo at myelin.hut.fi Tue Apr 28 07:57:08 1998 From: aapo at myelin.hut.fi (Aapo Hyvarinen) Date: Tue, 28 Apr 1998 14:57:08 +0300 Subject: ICA'99 call for papers Message-ID: <199804281157.OAA06512@myelin.hut.fi> -- We apologize if you receive multiple copies of this message. First Call for Papers: ------------- I C A ' 9 9 ------------- International Workshop on INDEPENDENT COMPONENT ANALYSIS and BLIND SIGNAL SEPARATION January 11-15, 1999 Aussois, France http://sig.enst.fr/~ica99 Submission deadline: July 15, 1998 ---------------------------------------------------------------------------- SCOPE ---------------------------------------------------------------------------- The workshop is devoted to recent advances in Independent Component Analysis and Blind Separation of Signals. It is intended to bring together researchers from the fields of artificial neural networks, signal processing, statistics, data analysis and all other domains connected to information processing. We are soliciting contributions covering all aspects of ICA and BSS: theory, methods, implementation and recent experimental results. The workshop will feature poster and oral presentations (no parallel sessions) and many opportunities for exchanges and informal discussions. ---------------------------------------------------------------------------- SPECIAL SESSIONS ---------------------------------------------------------------------------- Three special sessions will be organized on ICA applications: T. Sejnowski, Salk Institute, USA: Biomedical applications K. Torkkola, Motorola, Phoenix, USA: Speech and audio applications Y. Deville, UPS, Toulouse, France: General applications of ICA and BSS ---------------------------------------------------------------------------- VENUE ---------------------------------------------------------------------------- This one-week workshop will be held in Aussois, a small ski resort (alt. 1500m) in the magnificent mountains of La Vanoise in the heart of the French Alpes. The venue offers all the workshop facilities and we expect an enjoyable and productive `workshop atmosphere'. ---------------------------------------------------------------------------- SUBMISSION and PUBLICATION ---------------------------------------------------------------------------- Submission information will be available from our web site: http://sig.enst.fr/~ica99 Important dates: July 15, 1998 Submission of *full* paper Sep. 30, 1998 Notification of acceptance Jan. 11-15, 1999 Workshop All papers presented at the workshop will be collected in a volume of proceedings, which will be distributed to the participants on site. ---------------------------------------------------------------------------- SCIENTIFIC COMMITTEE ---------------------------------------------------------------------------- Shun-ichi Amari Brain Science Institute, RIKEN, Japan Tony Bell Salk Institute, USA Andrzej Cichocki Brain Science Institute, RIKEN, Japan Pierre Comon Eurecom, France Gustave Deco Siemens Research, Germany. Lieven De Lathauwer Katholieke Universiteit Leuven, Belgium Colin Fyfe University of Paisley, UK Simon Godsill University of Cambridge, UK Jean-Louis Lacoume Institut Nat. Polytechnique de Grenoble, France Ruey-Wen Liu University of Notre Dame, USA Odile Macchi CNRS/LSS, France Jean-Pierre Nadal Ecole Normale Superieure, Paris, France Erkki Oja Helsinki University of Technology, Finland Dinh-Tuan Pham CNRS/IMAG, France Jitendra Tugnait Auburn University, USA ---------------------------------------------------------------------------- ORGANIZATION ---------------------------------------------------------------------------- Organizers: Jean-Francois Cardoso CNRS and ENST, Paris, France Christian Jutten Institut Nat. Polytechnique de Grenoble, France Philippe Loubaton Universite de la Marne la Vallee, France Publicity: Aapo Hyvarinen Helsinki University of Technology, Finland Lieven De Lathauwer Katholieke Universiteit Leuven, Belgium ---------------------------------------------------------------------------- CONTACT INFORMATION ---------------------------------------------------------------------------- For more information contact: Jean-Francois Cardoso, ENST/SIG, 46 rue Barrault F-75634 Paris Cedex 13, France Internet: web site http://sig.enst.fr/~ica99 email ica99 at sig.enst.fr ---------------------------------------------------------------------------- From oby at cs.tu-berlin.de Wed Apr 29 06:37:47 1998 From: oby at cs.tu-berlin.de (Klaus Obermayer) Date: Wed, 29 Apr 1998 12:37:47 +0200 (MET DST) Subject: preprints available Message-ID: <199804291037.MAA16314@pollux.cs.tu-berlin.de> Dear Connectionists, I am happy to announce a series of papers on topographic clustering, self-organizing maps, dissimilarity data, and kernels. Cheers Klaus ------------------------------------------------------------------------ Prof. Klaus Obermayer phone: 49-30-314-73442 FR2-1, NI, Informatik 49-30-314-73120 Technische Universitaet Berlin fax: 49-30-314-73121 Franklinstrasse 28/29 e-mail: oby at cs.tu-berlin.de 10587 Berlin, Germany http://ni.cs.tu-berlin.de/ ========================================================================= A Stochastic Self-organizing Map for Proximity Data T. Graepel and K. Obermayer We derive an efficient algorithm for topographic mapping of proximity data (TMP), which can be seen as an extension of Kohonen's Self- Organizing Map to arbitrary distance measures. The TMP cost function is derived in a Baysian framework of Folded Markov Chains for the description of autoencoders. It incorporates the data via a dissimilarity matrix ${\mathcal D}$ and the topographic neighborhood via a matrix ${\mathcal H}$ of transition probabilities. From the principle of Maximum Entropy a non-factorizing Gibbs-distribution is obtained, which is approximated in a mean-field fashion. This allows for Maximum Likelihood estimation using an EM-algorithm. In analogy to the transition from Topographic Vector Quantization (TVQ) to the Self-organizing Map (SOM) we suggest an approximation to TMP which is computationally more efficient. In order to prevent convergence to local minima, an annealing scheme in the temperature parameter is introduced, for which the critical temperature of the first phase-transition is calculated in terms of ${\mathcal D}$ and ${\mathcal H}$. Numerical results demonstrate the working of the algorithm and confirm the analytical results. Finally, the algorithm is used to generate a connection map of areas of the cat's cerebral cortex. to appear in: Neural Computation preprint: http://ni.cs.tu-berlin.de/publications/#journals ------------------------------------------------------------------------- Fuzzy Topographic Kernel Clustering} T. Graepel and K. Obermayer A new topographic clustering algorithm is proposed, which - by the use of integral operator kernel functions - efficiently estimates the centers of clusters in high-dimensional feature spaces, which is related to data space by some nonlinear map. Like in the Self-Organizing Map topography is imposed by assuming finite transition probabilities between cluster indices. The optimization of the associated cost function is achieved by estimating the parameters via an EM-scheme and deterministic annealing. The effect of different radial basis function kernels on topographic maps of handwritten digit data is examined in computer simulations. In: W. Brauer, editor, Proceedings of the 5th GI Workshop Fuzzy Neuro Systems '98, pages 90-97, 1998. preprint: http://ni.cs.tu-berlin.de/publications/#conference ------------------------------------------------------------------------- An Annealed Self-Organizing Map for Source-Channel Coding M. Burger, T. Graepel, and K. Obermayer We derive and analyse robust optimization schemes for noisy vector quantization on the basis of deterministic annealing. Starting from a cost function for central clustering that incorporates distortions from channel noise we develop a soft topographic vector quantization algorithm (STVQ) which is based on the maximum entropy principle and which performs a maximum-likelihood estimate in an expectation-maximization (EM) fashion. Annealing in the temperature paramete $\beta$ leads to phase transitions in the existing code vector representation during the cooling process for which we calculate critical temperatures and modes as a function of eigenvectors and eigenvalues of the covariance matrix of the data and the transition matrix of the channel noise. A whole family of vector quantization algorithms is derived from STVQ, among them a deterministic annealing scheme for Kohonen's self-organizing map (SOM). This algorithm, which we call SSOM, is then applied to vector quantization of image data to be sent via a noisy binary symmetric channel. The algorithm's performance is compared to those of LBG and STVQ. While it is naturally superior to LBG, which does not take into account channel noise, its results compare very well to those of STVQ, which is computationally much more demanding. to appear in: NIPS 10 proceedings preprint: http://ni.cs.tu-berlin.de/publications/#conference The theory is quite well described in: T. Graepel, M. Burger, and K. Obermayer. Phase transitions in Stochastic Self-Organizing Maps. Phys. Rev. E, 56(4):3876-3890, 1997. preprint: http://ni.cs.tu-berlin.de/publications/#journals ----------------------------------------------------------------------- Review-style paper for the practioneers: Self-Organizing Maps: Generalization and New Optimization Techniques T. Graepel, M. Burger, and K. Obermayer We offer three algorithms for the generation of topographic mappings to the practitioner of unsupervised data analysis. The algorithms are each based on the minimization of a cost function which is performed using an EM algorithm and deterministic annealing. The soft topographic vector quantization algorithm (STVQ) -- like the original Self-Organizing Map (SOM) -- provides a tool for the creation of self-organizing maps of Euclidean data. Its optimization scheme, however, offers an alternative to the heuristic stepwise shrinking of the neighborhood width in the SOM and makes it possible to use a fixed neighborhood function solely to encode desired neighborhood relations between nodes. The kernel-based soft topographic mapping (STMK) is a generalization of STVQ and introduces new distance measures in data space based on kernel functions. Using the new distance measures corresponds to performing the STVQ in a high-dimensional feature space, which is related to data space by a nonlinear mapping. This preprocessing can reveal structure of the data which may go unnoticed if the STVQ is performed in the standard Euclidean space. The soft topographic mapping for proximity data (STMP) is another generalization of STVQ that enables the user to generate topographic maps for data which are given in terms of pairwise proximities. It thus offers a flexible alternative to multidimensional scaling methods and opens up a new range of applications for Self-Organizing Maps. Both STMK and STMP share the robust optimization properties of STVQ due to the application of deterministic annealing. In our contribution we discuss the algorithms together with their implementation and provide detailed pseudo-code and explanations. to appear in: Neurocomputing preprint: http://ni.cs.tu-berlin.de/publications/#journals From annesp at vaxsa.csied.unisa.it Wed Apr 29 07:05:10 1998 From: annesp at vaxsa.csied.unisa.it (annesp@vaxsa.csied.unisa.it) Date: Wed, 29 Apr 1998 12:05:10 +0100 Subject: E.R.CAIANIELLO SUMMER SCHOOL DEADLINE Message-ID: <98042912051019@vaxsa.csied.unisa.it> ***************************************************************** Please post **************************************************************** International Summer School ``Neural Nets E. R. Caianiello" 3rd Course "A Course on Speech Processing, Recognition, and Artificial Neural Networks" web page: http://wsfalco.ing.uniroma1.it/Speeschool.html The school is jointly organized by: INTERNATIONAL INSTITUTE FOR ADVANCED SCIENTIFIC STUDIES (IIASS) Vietri sul Mare (SA) Italy, ETTORE MAJORANA FOUNDATION AND CENTER FOR SCIENTIFIC CULTURE (EMFCSC) Erice (TR), Italy Supported by: EUROPEAN SPEECH COMMUNICATION ASSOCIATION (ESCA) Sponsored by: SALERNO UNIVERSITY, Dipartimento di Scienze Fisiche E.R. Caianiello (Italy) DIRECTORS OF THE COURSE DIRECTORS OF THE SCHOOL AND ORGANIZING COMMITTEE: Gerard Chollet (France). Maria Marinaro (Italy) M. Gabriella Di Benedetto (Italy) Michael Jordan (USA) Anna Esposito (Italy) Maria Marinaro (Italy) PLACE: International Institute for Advanced Scientific Studies (IIASS) Via Pellegrino 19, 84019 Vietri sul Mare, Salerno (Italy) DATES: 5th-14th October 1998 POETIC TOUCH Vietri (from "Veteri", its ancient Roman name) sul Mare ("on sea") is located within walking distance from Salerno and marks the beginning of the Amalfi coast. Short rides take to Positano, Sorrento, Pompei, Herculaneum, Paestum, Vesuvius, or by boat, the islands of Capri, Ischia, and Procida. Velia (the ancient "Elea" of Zeno and Parmenide) is a hundred kilometers farther down along the coast. Student Fee: 1500 dollars Student fee include accommodations (arranged by the school), meals, one day of excursion, and a copy of the proceedings of the school. Transportation is not included. A few scholarships are available for students who are otherwise unable to participate at the school, and who cannot apply for the grants offered by ESCA. The scholarship will partially cover lodging and living expenses. Day time: 3 hour in the morning, three hour in the afternoon. Day free: One day with an excursion of the places around. AIMS: The aim of this school is to present the experiments, the theories and the perspectives of acoustic phonetics, as well as to discuss recent results in the speech literature. The school aims to provide a background for further study in many of the fields related to speech science and linguistics, including automatic speech recognition. The school will bring together leading researchers and selected students in the field of speech science and technology to discuss and disseminate the latest techniques. The school is devoted to an international audience and in particular to all students and scientists who are working on some aspects of speech and want to learn other aspects of this discipline. MAJOR TOPICS The school will cover a number of broad themes relevant to speech, among them: 1) Speech production and acoustic phonetics 2) Articulatory, acoustic, and prosodic features 3) Acoustic cues in speech perception 4) Models of speech perception 5) Speech processing (Preprocessing algorithms for Speech) 6) Neural Networks for automatic speech recognition 7) Multi-modal speech recognition and recognition in adverse environments. 8) Speech to speech translation (Vermobil and CSTAR projects) 9) Applications (Foreign Language training aids, aids for handicapped, ....). 10) Stochastic Models and Dialogue systems FORMAT The meeting will follow the usual format of tutorials and panel discussions together with poster sessions for contributed papers. The following tutorials are planned: ABEER ALWAN UCLA University (CA) USA "Models of Speech Production and Their Application in Coding and Recognition" ANDREA CALABRESE University of Connecticut (USA) "Prosodic and Phonological Aspects of Language" GERARD CHOLLET CNRS - ENST France "ALISP, Speaker Verification, Interactive Voice Servers" PIERO COSI CNR-Padova Italy "Auditory Modeling and Neural Networks" RENATO DE MORI Universite d' Avignon, France "Statistical Methods for Automatic Speech Recognition" M. GABRIELLA DI BENEDETTO Universita' degli Studi di Roma "La Sapienza", Rome, Italy ``Acoustic Analysis and Perception of Classes of Sounds (vowels and consonants)" BJORN GRANSTROM Royal Institute of Technology (KTH) Sweden "Multi-modal Speech Synthesis with Application" JEAN P. HATON Universite Henri-Poincare, CRIN-INRIA, France "Neural Networks for Automatic Speech Recognition" HYNEK HERMANSKY Oregon Graduate Institute, USA "Goals and Techniques of Speech Analysis" HERMANN NEY Computer Science Department, Aachen Germany "Algorithms for Large Vocabulary Speech Recognition" "Text to Speech Translation using Statistical Methods" JOHN OHALA University of California at Berkeley (CA) USA "Articulatory Constraints on Distinctive Features" JEAN SYLVAIN LIENARD LIMSI-CNRS, France "Speech Perception, Voice Perception" "Beyond Pattern Recognition" PROCEEDINGS The proceedings will be published in the form of a book containing tutorial chapters written by the lecturers and possibly shorter papers from other participants. One free copy of the book will be distributed to each participant. LANGUAGE The official language of the school will be English. POSTER SUBMISSION There will be a poster session for contributed presentations from participants. Proposals consisting of a one page abstract for review by the organizers should be submitted with applications. DURATION Participants are expected to arrive in time for the evening meal on Sunday 4th October and depart on Tuesday 15th October. Sessions will take place from Monday 5th-Wednesday 14th. COST The cost per participant of 1.500 $ dollars covers accommodation (in twin rooms), meals for the duration of the course, and one day of excursion. -- A supplement of 40 dollars per night should be paid for single room. Payment details will be notified with acceptance of applications. GRANTS -- A few ESCA grants are available for participants (which cover tuition and, maybe, part of the lodging). See http://ophale.icp.inpg.fr/esca/grants.html for further information. Individual applications for grants should be sent to Wolfgang Hess by e-mail: wgh at sunwgh.ikp.uni-bonn.de ELIGIBILITY The school is open to all suitably qualified scientists from around the world. APPLICATION PROCEDURE: Important Date: Application deadline: May 15 1998 Notification of acceptance: May 30 1998 Registration fee payment deadline: July 10 1998 People with few years of experience in the field should include a recommendation letter of their supervisor or group leader Places are limited to a maximum of 60 participants in addition to the lecturers. These will be allocated on a first come, first served basis. ************************************************************************** APPLICATION FORM Title:^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Family Name:^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Other Names:^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Name to appear on badge:^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Mailing Address (include institution or company name if appropriate): ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Phone:^^^^^^^^^^^^^^^^^^^^^^Fax:^^^^^^^^^^^^^^^^^^^ E-mail:^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Date of Arrival : Date of Departure: Will you be applying for a ESCA grant ? yes/no* *(please delete the alternatives which do not apply) Will you be applying for a scholarship ? yes/no* *(please delete the alternatives which do not apply) *(please include in your application a justification for scholarship request) ***************************************************************** Please send the application form together the recommendation letter by electronic mail to: iiass at tin.it, subject: summer school; or by fax: +39 89 761 189 (att.ne Prof. M. Marinaro) or by ordinary mail to the address below: IIASS Via Pellegrino 19, I84019 Vietri sul Mare (Sa) Italy For further information please contact: Anna Esposito International Institute for advanced Scientific Studies (IIASS) Via Pellegrino, 19, 84019 Vietri sul Mare (SA) Italy Fax: + 39 89 761189 e-mail: annesp at vaxsa.csied.unisa.it ================== RFC 822 Headers ================== From wjf at eng.cam.ac.uk Wed Apr 1 03:58:02 1998 From: wjf at eng.cam.ac.uk (W J Fitzgerald) Date: Wed, 01 Apr 1998 08:58:02 +0000 Subject: BAYESIAN SIGNAL PROCESSING Message-ID: <117594.3100409882@wjfmac.eng.cam.ac.uk> Dear Friends, I hope this might be of interest to some of you. regards Bill Fitzgerald Isaac Newton Institute for Mathematical Sciences EC SUMMER SCHOOL BAYESIAN SIGNAL PROCESSING 19 - 31 July 1998 Organisers: WJ Fitzgerald (Cambridge), RL Smith (North Carolina), AT Walden (Imperial), PC Young (Lancaster) The main focus and thrust of this workshop will be that Bayesian methods provide a unifying methodology whereby different kinds of mathematical models may be examined within a common statistical framework. The workshop will bring together the statistical and computational expertise of leading statisticians and the modelling expertise of mathematicians and subject matter specialists, with the broad objective of developing new signal processing tools which make efficient use of modern computational resources while combining the most up-to-date research of both groups of specialists. Specific topics to be covered include: - Bayesian methods in general and numerical methods in particular; - Nonlinear and nonstationary time series estimation; - Forecasting and changepoint modelling; - Nonlinear signal processing in econometrics and financial time series; - Dynamical systems and statistics; - Environmental applications and spatial data analysis. The workshop will take place at the start of a six month programme on Nonlinear and Nonstationary Signal Processing to be held at the Isaac Newton Institute in Cambridge (July - Dec 1998), where it is hoped that many of the problems identified during the workshop will be studied in detail by the participants. Grants: The conference is supported by a grant from the European Community which will provide funding towards the registration, travel and subsistence costs of selected young (under 35 years) participants. Applications from women and anyone living in Greece, Ireland and Portugal and other less favoured regions of the European Community are particularly encouraged. Other limited funds exist for participants from outside the EC. Self-supporting applications of any age and nationality are welcome. Applications: The workshop will take place at the Newton Institute and accommodation for participants will be provided at Christ's College. The conference package costs ?650, which includes registration fees, accommodation, breakfast and evening meals plus lunch and refreshments during the days that lectures take place. Further Information and Application Forms: are available from the WWW at http://www.newton.cam.ac.uk/programs/nspw01.html Completed application forms should be sent to Heather Hughes at the Newton Institute, or via email to h.hughes at newton.cam.ac.uk The programme home page is at http://www.newton.cam.ac.uk/programs/nsp.html **Closing Date for the receipt of applications is 24 April 1998** ******************************************************************** * Dr. W.J.Fitzgerald, http://www-sigproc.eng.cam.ac.uk/~wjf * * Signal Processing Group, Cambridge University Engineering Dept. * * Cambridge CB2 1PZ, U.K. * * and/or * * Christ's College, * * Cambridge CB2 3BU * * * * tel: +44-(1223)-332719, fax: +44-(1223)-332662 * * email: wjf at eng.cam.ac.uk * ******************************************************************** From zemel at U.Arizona.EDU Thu Apr 2 18:40:26 1998 From: zemel at U.Arizona.EDU (Richard Zemel) Date: Thu, 2 Apr 1998 16:40:26 -0700 (MST) Subject: postdoctoral position Message-ID: POSTDOCTORAL RESEARCH ASSOCIATE UNIVERSITY OF ARIZONA Tucson, Arizona A postdoctoral position is available in the laboratory of Richard Zemel. It is expected that the applicant will collaborate in research projects in at least one of these areas: (1) theoretical models of neural coding; (2) studies of spatial and object attention; (3) motion processing in natural and artificial systems; (4) cue combination. For more details, see http://www.u.arizona.edu/~zemel. Numerous opportunities exist for interactions and collaborations with other researchers at the University of Arizona. The university has a strong interdisciplinary group in the area of Cognition and Neural Systems, including active research in: neural coding and spatial cognition (Carol Barnes, Bruce McNaughton, Lynn Nadel); object perception and attention (Mary Peterson); visual neurophysiology (Peter DeWeerd, Fraser Wilson); multimodal integration (Felice Bedford, Kerry Green) and others. For more information see http://w3.arizona.edu/~psych/cns.htm. Applicants should have a strong background and education in a quantitative discipline, such as physics, mathematics, statistics, or computer science. Knowledge of neuroscience or psychophysics is also desirable. Starting date is flexible, ranging from immediately to September. The position is available for 1-2 years depending on accomplishment. Salary is competitive. Please send a CV, a letter describing research interests and background, and at least 2 letters of recommendation by post or email to: Richard Zemel Department of Psychology University of Arizona Tucson, AZ 85721 Email: zemel at u.arizona.edu From becker at curie.psychology.mcmaster.ca Thu Apr 2 20:31:56 1998 From: becker at curie.psychology.mcmaster.ca (Sue Becker) Date: Thu, 2 Apr 1998 20:31:56 -0500 (EST) Subject: COMPUTATIONAL NEUROSCIENCE POSTDOC Message-ID: COMPUTATIONAL NEUROSCIENCE POSTDOCTORAL POSITION AVAILABLE Department of Psychology McMaster University A postdoctoral position is open in the Psychology Department of McMaster University, Hamilton, Ontario. A multidisciplinary approach will be taken to develop biologically plausible models of hippocampal and neocortical memory systems. Projects will include developing simulations of cortical-cortical and subcortical-cortical interactions during learning and information storage. In addition to our focus on processing in hippocampal-cortical systems, we are also investigating and modelling the role of cortico-thalamic back-projections. Research projects will be conducted in close collaboration with R. Racine, a neuroscientist, S. Becker, a computational modeller, and S. Haykin, an electrical engineer. The primary goal of this collaborative effort is to build powerful learning algorithms in neural networks which are based on rules suggested by both memory research and physiology research (e.g. LTP work). Racine's laboratory has recently provided the first demonstrations of LTP in the neocortex of the awake, freely-moving rat. The rules that apply to LTP induction in neocortical systems are quite different from those determined for the hippocampus. An OPTIONAL component of this postdoctoral position would be participation in further experimental investigations of neocortical LTP in either slice or in vivo preparations. This project is funded by a collaborative research grant from the Natural Sciences and Engineering Research Council of Canada to R. Racine, S. Haykin and S. Becker. Please send curriculum vitae, expression of interest, and the names and e-mail or phone numbers of three references to Ron Racine at racine at mcmail.cis.mcmaster.ca From jb at uran.informatik.uni-bonn.de Fri Apr 3 08:37:19 1998 From: jb at uran.informatik.uni-bonn.de (Joachim M. Buhmann) Date: Fri, 03 Apr 1998 15:37:19 +0200 Subject: Tech. Report on Unsupervised Learning Message-ID: <1.5.4.32.19980403133719.00665340@uran.informatik.uni-bonn.de> The following technical report on Unsupervised Learning is now available from our website http://www-dbv.informatik.uni-bonn.de/papers.html#NeuralNetworks Empirical Risk Approximation: An Induction Principle for Unsupervised Learning Joachim M. Buhmann Institute for Computer Science (III), University of Bonn Unsupervised learning algorithms are designed to extract structure from data without reference to explicit teacher information. The quality of the learned structure is determined by a cost function which guides the learning process. This paper proposes Empirical Risk Approximation as a new induction principle for unsupervised learning. The complexity of the unsupervised learning models are automatically controlled by the two conditions for learning: (i) the empirical risk of learning should uniformly converge towards the expected risk; (ii) the hypothesis class should retain a minimal variety for consistent inference. The maximal entropy principle with deterministic annealing as an efficient search strategy arises from the Empirical Risk Approximation principle as the optimal inference strategy for large learning problems. Parameter selection of learnable data structures is demonstrated for the case of K-means clustering. --------------------------------------------------------------------- Joachim M. Buhmann Institut fuer Informatik III Tel.(office) : +49 228 734 380 Universitaet Bonn Tel.(secret.): +49 228 734 292 Roemerstr. 164 Fax: +49 228 734 382 D-53117 Bonn email: jb at informatik.uni-bonn.de Fed. Rep. Germany jb at cs.bonn.edu http://www-dbv.informatik.uni-bonn.de --------------------------------------------------------------------- From wahba at stat.wisc.edu Fri Apr 3 17:24:19 1998 From: wahba at stat.wisc.edu (Grace Wahba) Date: Fri, 3 Apr 1998 16:24:19 -0600 (CST) Subject: Model Selection, RanTraceGACV Message-ID: <199804032224.QAA02784@hera.stat.wisc.edu> Im taking the liberty of noting that the following paper will be of interest to those interested in model selection: Jianming Ye, `On measuring anc correcting the effects of data mining and model selection', J. Amer. Statist. Assoc. 93, March 1998, pp 120- 131. ......... The following short abstract discusses the randomized trace technique for computing the GACV (Generalized Approximate Cross Validation function) for Bernoulli Data which was proposed in Xiang and Wahba, Statistica Sinica 1996, 675-692. Xiang, D. and Wahba, G. `Approximate Smoothing Spline Methods for Large Data Sets in the Binary Case', TR 982. http://www.stat.wisc.edu/~wahba -> TRLIST From baluja at jprc.com Mon Apr 6 15:18:01 1998 From: baluja at jprc.com (Shumeet Baluja) Date: Mon, 6 Apr 1998 15:18:01 -0400 Subject: paper available: Fast Probabilistic Modeling for Combinatorial Optimization Message-ID: <199804061918.PAA01197@india.jprc.com> Fast Probabilistic Modeling for Combinatorial Optimization ----------------------------------------------------------- Shumeet Baluja Justsystem Pittsburgh Research Center & Carnegie Mellon University Scott Davies Carnegie Mellon University Abstract Probabilistic models have recently been utilized for the optimization of large combinatorial search problems. However, complex probabilistic models that attempt to capture inter-parameter dependencies can have prohibitive computational costs. The algorithm presented in this paper, termed COMIT, provides a method for using probabilistic models in conjunction with fast search techniques. We show how COMIT can be used with two very different fast search algorithms: hillclimbing and Population-based incremental learning (PBIL). The resulting algorithms maintain many of the benefits of probabilistic modeling, with far less computational expense. Extensive empirical results are provided; COMIT has been successfully applied to jobshop scheduling, traveling salesman, and knapsack problems. This paper also presents a review of probabilistic modeling for combinatorial optimization. This paper is an extension of the earlier work reported in: Combining Multiple Optimization Runs with Optimal Dependency Trees by: S. Baluja & S. Davies, June 1997. Available from: http://www.cs.cmu.edu/~baluja Questions and comments are welcome. Shumeet Baluja From aapo at myelin.hut.fi Tue Apr 7 12:32:32 1998 From: aapo at myelin.hut.fi (Aapo Hyvarinen) Date: Tue, 7 Apr 1998 19:32:32 +0300 Subject: FastICA package for MATLAB Message-ID: <199804071632.TAA26001@myelin.hut.fi> FastICA, a new MATLAB package for independent component analysis, is now available at: http://www.cis.hut.fi/projects/ica/fastica/ FastICA is a public-domain package that implements the fast fixed-point algorithm for ICA, and features an easy-to-use graphical user interface. The fixed-point algorithm is a computationally highly efficient method for ICA: in independent experiments it has been found to be 10-100 times faster than conventional gradient descent methods for ICA. Another advantage of the fixed-point algorithm is that it can be used to perform projection pursuit, estimating the independent components one-by-one. Aapo Hyvarinen on behalf of the FastICA Team at the Helsinki University of Technology fastica at mail.cis.hut.fi From marco at idsia.ch Tue Apr 7 08:58:29 1998 From: marco at idsia.ch (Marco Wiering) Date: Tue, 7 Apr 1998 14:58:29 +0200 Subject: Paper announcement !!!!!!! Message-ID: <199804071258.OAA03950@ruebe.idsia.ch> Fast Online Q(lambda) Marco Wiering Juergen Schmidhuber To appear in the Machine Learning Journal Q(lambda)-learning uses TD(lambda)-methods to accelerate Q-learning. The update complexity of previous online Q(lambda) implementations based on lookup-tables is bounded by the size of the state/action space. Our faster algorithm's update complexity is bounded by the number of actions. The method is based on the observation that Q-value updates may be postponed until they are needed. Also to be presented at the 10th European Conference On Machine Learning (ECML'98), Chemnitz (Germany), April 21-24 1998. FTP-host: ftp.idsia.ch FTP-files: /pub/marco/fast_q.ps.gz /pub/marco/ecml_q.ps.gz WWW: http://www.idsia.ch/~marco/publications.html http://www.idsia.ch/~juergen/onlinepub.html Marco & Juergen IDSIA, Switzerland From todd at cs.ua.edu Thu Apr 9 16:41:16 1998 From: todd at cs.ua.edu (Todd Peterson) Date: Thu, 9 Apr 1998 15:41:16 -0500 (CDT) Subject: Paper Available Message-ID: <199804092041.PAA02777@todd.cs.ua.edu> Paper available: An RBF Network Alternative for a Hybrid Architecture Todd Peterson, Ron Sun Department of Computer Science The University of Alabama Tuscaloosa, AL 35487 EMAIL: todd,rsun at cs.ua.edu ABSTRACT Although a previous model CLARION has shown some measure of success in sequential decision making tasks by utilizing a hybrid architecture that uses both procedural and declarative learning, it had some problems resulting from the use of backpropagation networks. CLARION-RBF is a more parsimonious architecture that remedies some of the problems exhibited in CLARION, by utilizing RBF Networks. CLARION-RBF is capable of learning reactive procedures as well as having high level symbolic knowledge extracted and applied. To appear in IJCNN, Anchorage, AK, May 4-9, 1998. The Postscript version is accessible through: http://cs.ua.edu/~rsun/tp.rbf.ps or from: http://cs.ua.edu/~todd/ijc.ps From sknerr at ireste.fr Fri Apr 10 12:06:44 1998 From: sknerr at ireste.fr (stefan knerr) Date: Fri, 10 Apr 1998 18:06:44 +0200 Subject: PhD Thesis Message-ID: <352E4394.3218@ireste.fr> PhD Studentship in Document Image Processing Department of Image and Video Processing University Nantes - IRESTE, France DOCUMENT IMAGE MODELING AND SEGMENTATION USING STATISTICAL 2-D APPROACHES Early processing stages of document analysis systems such as the localization and segmentation of text fields, logos, graphics, lines, etc. usually rely on heuristic methods or must be defined by hand. The goal of this thesis is to elaborate a principled approach to these problems based on statistical approaches such as Markov Random Fields and 2-D Hidden Markov Models. This is an exciting opportunity to work on a real world problem within an environment which includes an industrial partner as well as a University research lab. Several possibilities for financing a 3 year PhD work exist depending on the qualification and on the nationality of the candidate. Interested candidates should send a CV, related publications or reports, names and addresses of one or two persons (professor, lecturer) or companies who will recommend you to: Stefan Knerr (sknerr at ireste.fr) or Christian Viard-Gaudin (cviard at ireste.fr) IRESTE Rue Christian Pauc, La Chantrerie, BP 60601 44306 Nantes Cedex 3 France Applications should arrive before Mai 31, 1998. -- Stefan Knerr Laboratoire SEI, IRESTE Rue Christian Pauc, La Chantrerie, BP 60601 44306 Nantes Cedex 3, France Tel. +33 2 40683036 Fax. +33 2 40683066 sknerr at ireste.fr From ingber at ingber.com Mon Apr 13 13:54:19 1998 From: ingber at ingber.com (Lester Ingber) Date: Mon, 13 Apr 1998 12:54:19 -0500 Subject: Paper: Volatility Volatility of Financial Markets Message-ID: <19980413125419.A3006@ingber.com> The paper markets98_vol.ps.Z [120K] is available at my InterNet archive: %A L. Ingber %A J.K. Wilson %T Volatility of volatility of financial markets %R DRW-98-1-VVFM %I DRW Investments LLC %C Chicago, IL %D 1998 %O URL http://www.ingber.com/markets98_vol.ps.Z We present empirical evidence for considering volatility of Eurodollar futures as a stochastic process, requiring a generalization of the standard Black-Scholes (BS) model which treats volatility as a constant. We use a previous development of a statistical mechanics of financial markets (SMFM) to model these issues. ======================================================================== Instructions for Retrieval of Code and Reprints Interactively Via WWW The archive can be accessed via WWW path http://www.ingber.com/ http://www.alumni.caltech.edu/~ingber/ where the last address is a mirror homepage for the full archive. Interactively Via Anonymous FTP Code and reprints can be retrieved via anonymous ftp from ftp.ingber.com. Interactively [brackets signify machine prompts]: [your_machine%] ftp ftp.ingber.com [Name (...):] anonymous [Password:] your_e-mail_address [ftp>] binary [ftp>] ls [ftp>] get file_of_interest [ftp>] quit The 00index file contains an index of the other files. Files have the same WWW and FTP paths under the main / directory; e.g., http://www.ingber.com/MISC.DIR/00index_misc and ftp://ftp.ingber.com/MISC.DIR/00index_misc reference the same file. Electronic Mail If you do not have WWW or FTP access, get the Guide to Offline Internet Access, returned by sending an e-mail to mail-server at rtfm.mit.edu with only the words send usenet/news.answers/internet-services/access-via-email in the body of the message. The guide gives information on using e-mail to access just about all InterNet information and documents. Additional Information Limited help assisting people with queries on my codes and papers is available only by electronic mail correspondence. Sorry, I cannot mail out hardcopies of code or papers. Lester ======================================================================== -- /* Lester Ingber Lester Ingber Research * * ingber at ingber.com http://www.ingber.com/ ftp.ingber.com * * ingber at alumni.caltech.edu http://www.alumni.caltech.edu/~ingber/ * * PO Box 06440 Wacker Dr PO - Sears Tower Chicago, IL 60606-0440 */ From aperez at lslsun.epfl.ch Tue Apr 14 04:12:41 1998 From: aperez at lslsun.epfl.ch (Andres Perez-Uribe) Date: Tue, 14 Apr 1998 10:12:41 +0200 Subject: New Book: Bio-Inspired Computing Machines Message-ID: <35331A79.E99B3A3D@lslsun.epfl.ch> Dear Connectionist, This is to announce a new book entitled: "Bio-Inspired Computing Machines Toward Novel Computational Architectures" Daniel Mange and Marco Tomassini (Eds.) Presses polytechniques et universitaires romande, Lausanne, Switzerland http://lslwww.epfl.ch/pages/publications/books/1998_1/ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Originality: This book is unique for the following reasons: - It follows a unified approach to bio-inspiration based on the so-called POE model: phylogeny (evolution of species), ontogeny (development of individual organisms), and epigenesis (life-time learning). - It is largely self-contained, with an introduction to both biological mechanisms (POE) and digital hardware (digital systems, cellular automata). - It is mainly applied to computer hardware design. - It is largely self-contained, with an introduction to both biological mechanisms (POE) and digital hardware (digital systems, cellular automata). - It is mainly applied to computer hardware design. BACK-COVER TEXT This volume, written by experts in the field, gives a modern, rigorous and unified presentation of the application of biological concepts to the design of novel computing machines and algorithms. While science has as its fundamental goal the understanding of Nature, the engineering disciplines attempt to use this knowledge to the ultimate benefit of Mankind. Over the past few decades this gap has narrowed to some extent. A growing group of scientists has begun engineering artificial worlds to test and probe their theories, while engineers have turned to Nature, seeking inspiration in its workings to construct novel systems. The organization of living beings is a powerful source of ideas for computer scientists and engineers. This book studies the construction of machines and algorithms based on natural processes: biological evolution, which gives rise to genetic algorithms, cellular development, which leads to self-replicating and self-repairing machines, and the nervous system in living beings, which serves as the underlying motivation for artificial learning systems, such as neural networks. PUBLIC Undergraduate and graduate students, researchers, engineers, computer scientists, and communication specialists. TABLE OF CONTENTS Preface 1 An Introduction to Bio-Inspired Machines 2 An Introduction to Digital Systems 3 An Introduction to Cellular Automata 4 Evolutionary Algorithms and their Applications 5 Programming Cellular Machines by Cellular Programming 6 Multiplexer-Based Cells 7 Demultiplexer-Based Cells 8 Binary Decision Machine-Based Cells 9 Self-Repairing Molecules and Cells 10 L-hardware: Modeling and Implementing Cellular Development using L-systems 11 Artificial Neural Networks: Algorithms and Hardware Implementation 12 Evolution and Learning in Autonomous Robotic Agents Bibliography Index -- Andres PEREZ-URIBE Logic Systems Laboratory Computer Science Department Swiss Federal Institute of Technology-Lausanne 1015 Lausanne, Switzerland Email: aperez at lslsun.epfl.ch http://lslwww.epfl.ch/~aperez Tel: +41-21-693-2652 Fax: +41-21-693 3705 From harnad at coglit.soton.ac.uk Tue Apr 14 09:39:12 1998 From: harnad at coglit.soton.ac.uk (S.Harnad) Date: Tue, 14 Apr 1998 14:39:12 +0100 (BST) Subject: Connectionist Explanation: PSYC Call for Commentators (614 lines) Message-ID: <199804141339.OAA15840@amnesia.psy.soton.ac.uk> Green: CONNECTIONIST EXPLANATION The target article below has just appeared in PSYCOLOQUY, a refereed journal of Open Peer Commentary sponsored by the American Psychological Association. Qualified professional biobehavioral, neural or cognitive scientists are hereby invited to submit Open Peer Commentary on this article. Please email for Instructions if you are not familiar with PSYCOLOQUY format and acceptance criteria (all submissions are refereed). To submit articles and commentaries or seek information: EMAIL: psyc at pucc.princteton.edu URL: http://www.princeton.edu/~harnad/psyc.html http://www.cogsci.soton.ac.uk/psyc AUTHOR'S RATIONALE FOR SOLICITING COMMENTARY: My reason for soliciting commentary is quite straightforward. Connectionist models of cognition are quite common in the psychological literature these days, but there is very little discussion of the exact role they are thought to play in science. If they are indeed theories, in the traditional sense, then some explanation of the ways in which they seem to depart from traditional theories is needed. If they are not traditional theories, then a clear description of what they are, and an account of why we should pay attention to them, is needed. Such a discussion should take place among connectionists, philosophers of science and mind, and psychologists. Psycoloquy seems like an ideal vehicle for such a discussion. ----------------------------------------------------------------------- psycoloquy.98.9.04.connectionist-explanation.1.green Tue 14 Mar 1998 ISSN 1055-0143 (23 paragraphs, 16 references, 4 notes, 584 lines) PSYCOLOQUY is sponsored by the American Psychological Association (APA) Copyright 1998 Christopher D. Green ARE CONNECTIONIST MODELS THEORIES OF COGNITION? Christopher D. Green Department of Psychology York University North York, Ontario M3J 1P3 CANADA christo at yorku.ca http://www.yorku.ca/faculty/academic/christo ABSTRACT: This paper explores the question of whether connectionist models of cognition should be considered to be scientific theories of the cognitive domain. It is argued that in traditional scientific theories, there is a fairly close connection between the theoretical (unobservable) entities postulated and the empirical observations accounted for. In connectionist models, however, hundreds of theoretical terms are postulated -- viz., nodes and connections -- that are far removed from the observable phenomena. As a result, many of the features of any given connectionist model are relatively optional. This leads to the question of what, exactly, is learned about a cognitive domain modelled by a connectionist network. KEYWORDS: artificial intelligence, cognition, computer modelling, connectionism, epistemology, explanation, methodology, neural nets, philosophy of science, theory. 1. Connectionist models of cognition are all the rage now. It is not clear, however, in what sense such models are to be considered THEORIES of cognition. This may be problematic, for if connectionist models are NOT to be considered THEORIES of cognition, in the traditional scientific sense of the word, then the question arises as to what exactly they are, and why we should pay attention to them? If, on the other hand, they are to be regarded as scientific theories it should be possible to explicate precisely in what sense this is true, and to show how they fulfill the functions we normally associate with theories. In this paper, I begin by examining the question of what it is to be a scientific theory. Second, I describe in precisely what sense traditional computational models of cognition can be said to perform this role. Third, I examine whether or not connectionist models can be said to do the same. My conclusion is that connectionist models could, under a certain interpretation of what it is they model, be considered to be theories, but that this interpretation is likely to be unacceptable to many connectionists. 2. A typical complex scientific theory contains both empirical and theoretical terms. The empirical terms refer to observable entities. The theoretical terms refer to unobservable entities that improve the predictive power of the theory as a whole. The exact ontological status of objects referred to by theoretical terms is a matter of some debate. Realists believe them to be actual objects that resist direct observation for one reason or another. Instrumentalists consider them to be mere "convenient fictions" that earn their scientific keep merely by the predictive accuracy they lend to the theory. I think it is fair to say that the vast majority of research psychologists are realists about the theoretical terms they use, though they are, in the main, unreflective realists who have never seriously considered alternative possibilities. 3. Let us begin with a relatively uncontroversial theory from outside psychology -- Mendelian genetics. In the Mendelian scheme, entities called "genes" were said to be responsible for the propagation of traits from one generation of organisms to another. Mendel was unable to observe anything corresponding to "genes," but their invocation made it possible for him to predict correctly the proportions in which succeeding generations of organisms would express a variety of traits. As such, the gene is a classic example of a theoretical entity. For present purposes, it is important to note that each such theoretical gene, though unobservable, was hypothesized to correspond to an individual trait. That is, in addition to the predictive value each theoretical gene provided, each also justified its existence by being responsible for a particular phenomenon. There were no genes in the system that were not directly tied to the expression of a trait. Although some genes were said not to be expressed in the phenotype (viz., recessive genes in heterozygous individuals), all were said to be directly involved in the calculation of the expression of a specific trait. That is to say, their inclusion in the theory was justified in part by the SPECIFICITY of the role they were said to play. It is worth noting that the actual existence of genes remained controversial until the discovery of their molecular basis -- viz., DNA -- and our understanding of them changed considerably with that discovery. 4. Now consider, as a psychological example of theoretical entities, the model of memory proposed by Atkinson and Shiffrin (1971). It is a classic "box-and-arrow" theory. Information is fed from the sensory register into a holding space called Short Term Store (STS). If continuously rehearsed, a limited number of items can be stored there indefinitely. If the number of items exceeds the capacity of the store, some are lost. If rehearsal continues for an unspecified duration, it is claimed that some or all of these items are transferred to another holding space called Long Term Store (LTS). The capacity of LTS is effectively unlimited, and items in LTS need not be continuously rehearsed, but are said to be kept in storage effectively permanently. STS and LTS are, like genes, theoretical entities. They cannot be directly observed, but their postulation enables the psychologist to predict correctly a number of memory phenomena. In each such phenomenon, the activity of each store is carefully specified. The precision of this specification seems to be at least part of the reason that scientists are willing to accept them. Indeed, many experiments aimed at confirming their existence are explicitly designed to block, or interfere with the hypothesized activity of one in order to demonstrate the features of the "pure" activity of the other. Whether or not this could be successfully accomplished was once a dominant question in memory theory. The issue of short term memory EFFECTS being "contaminated" by the uncontrollable and unwanted activity of LTS occupied many experiments of the 1960s and 1970s. 5. Over the last 30 years the Atkinson and Shiffrin model has been elaborated and refined. As a result, the number of memory systems hypothesized to exist has grown tremendously. Baddeley (1992), for instance, has developed STS into a series of slave systems responsible for information entering memory from the various sense modalities (e.g., the phonological loop, the visuospatial sketchpad), the activities of which are coordinated by a central executive. Tulving (1985), on the other hand, has divided LTS into four hierarchically arranged systems responsible for episodic memory (for personal events), semantic memory (for general information), procedural memory (for skills) and implicit memory (for priming). In order to establish the existence of each of these many theoretical entities, thousands of experiments have been performed, aimed at revealing the independent of activity of one or another by attempts to block the activity of the others. 6. Once again, the question of whether the activity of a single memory system can be studied in isolation has called into question the very existence of that system. For over a decade, now, the elucidation of implicit memory phenomena has been a major issue in memory theory (Schacter, 1987, 1992; Roediger, 1990; Roediger & McDermott, 1993;). In the typical implicit memory experiment [1], subjects study a list of items (both words and pictures have been used) by processing them briefly. This can be as simple as reading the word or naming the object, or it can be more involved, such as deciding whether the items belongs to a certain class of items (e.g., is a car a kind vehicle?) or decomposing it in to parts (e.g., counting the number of letters in words, or counting the edges or corners in pictured items). The subjects then take part in a memory test, although they are not told that it is a memory test, and it could indeed be performed without having studied the material. In this test, they see a new list of items, some of which, unbeknownst to them, are the same as (or closely related to) the items they have studied. Such tests are sometimes puzzles of various sorts (e.g., completing incomplete words or identifying the items in incomplete pictures). Sometimes they are as simple as deciding whether the items are true words (as opposed to pronounceable non-words such as BLICK) or possible objects. People perform reliably better on these tasks when the items in question are ones that were on the study list (or closely related to items on the study list) than when the items are new. Upon post-experimental debriefing, however, they are often unable to say which items they had studied before and which they had not. This is the classic implicit memory effect. 7. Recently, however, it has been argued (Roediger & McDermott, 1993) that explicit memory may be "contaminating" the hypothesized effect of the implicit memory system. The degree of this contamination is not clear, but it is possible, in principle (though unlikely), that ALL implicit memory phenomena are the result of covert explicit memory. The evidence for this (Jacoby, 1991) comes from comparing the behavior of a typical implicit memory group with that of a control group that goes through the same procedure but is told EXPLICITLY that the answers to some of the test problems are items they studied before. The outcome is that these subjects do almost as well as the experimental subjects, thus calling into question the "implicitness" of the traditional subjects' memories. As a result, many have begun question the very existence of the implicit memory system. Many psychologists argue that the implicit memory effects are the result of a certain kind of processing of a more general memory system, not the autonomous activity of a distinct system of its own. 8. With the entry of computer models into psychology, the theories have become even more complex, using dozens of theoretical entities. A recent version of Chomskyan linguistic theory, for instance, postulates more than two dozen rules that are said to control the building and interpretation of grammatical sentences (see e.g., Berwick, 1985). But even here the empirical data must bear fairly directly on each theoretical entity. None of these rules is without specific predicted effects. Each of the rules performs a certain function without which the construction and interpretation of grammatical sentences could not proceed correctly. For example, RULE ATTACH-VP, sensibly enough, attaches verb phrases to sentences; RULE ATTACH-NOUN similarly attaches nouns to noun phrases; and so forth. Part of what justifies the inclusion in the theory of terms referring to each of these entities is the fact that they are explicitly connected to specific empirical phenomena. 9. In each of the models I have described so far, each theoretical entity represents something in particular, even if that something is itself theoretical. The existence and properties of the entities represented are supported by empirical evidence relating specifically to that entity. In a typical connectionist model, however, there are dozens, sometimes hundreds, of simple units, bound together by hundreds, sometimes thousands, of connections. Neither the units nor the connections represent anything known to exist in the cognitive domain the network is being used to model. Similarly, the rules that govern how the activity of one unit will affect the activity of other units to which it is connected are extremely simple, and not obviously related to the domain that the network is being used to model. Ditto for the rules that govern how the weights on the connections between units are to be changed. In particular, the units of the network are not thought to represent particular propositional attitudes (i.e., beliefs, desires, etc.) or the terms or concepts that might be thought to underlie them. This is all considered a distinct advantage among connectionists. Neither the units nor the connections correspond to anything in the way that variables and rules did in traditional computational models of cognition. Representations, to the degree that they are admitted at all, are said to be distributed across the activities of the units as a group. Any representation-level rules that the model is said to use are likewise distributed across the weights of all of the connections in the network. This gives connectionist networks their characteristic flexibility: they are able to learn in a wide variety of cognitive domains, to generalize their knowledge easily to new cases, to continue working reasonably well despite incomplete input or even moderate damage to their internal structure, etc. The only real question is whether they are, indeed, TOO flexible to be good theories. Or whether, by contrast, there are heretofore unrecognized features of good theories of which connectionist models can apprise us. 10. Each of the units, connections, and rules in a connectionist network is a theoretical entity. Each name referring to it in a description of the network is a theoretical term in the theory of cognition that it embodies [2]. With the previously described theories, it was evident that each theoretical entity had a specific job to do. If it were removed, not only would the performance of the model as a whole suffer, but it would suffer in predictable ways, viz., the particular feature of the model's performance for which the theoretical entity in question was responsible -- i.e., that which it represented -- would no longer obtain. The units and connections in a connectionist net -- precisely in virtue of the distributed nature of their activity -- need not bear any such relation to the various activities of the model. Although this seems to increase the model's overall efficiency, it also seems to undermine the justification for each of the units and connections in the network. To put things even more plainly, if one were to ask, say, of Berwick's (1985) symbolic model of grammar, "What is the justification for postulating RULE ATTACH-NOUN?" the answer would be quite straightforward: "Because without it nouns would not be attached to noun phrases and the resulting outputs would be ungrammatical." The answer to the parallel question with respect to the a connectionist network -- viz., "What is the justification for postulating (say) unit 123 in this network?" -- is not so straightforward. Precisely because connectionist networks are so flexible, the right answer is probably something like, "No reason in particular. The network would probably perform just as well without it" [3]. 11. If this is true, we are led to an even more pressing question: exactly what is it that we can actually be said to KNOW about a given cognitive process once we have modelled it with a connectionist network? In the case of, say, the Atkinson and Shiffrin model of memory, we can say that we have confirmation of the idea that there are at least two forms of memory store -- short and long term -- and this confirmation amounts to a justification of sorts for their postulation. Are we similarly to say that a particular connectionist model with, say, 326 units that correctly predicts activity in a given cognitive domain confirms the idea that there are exactly 326 units governing that activity? This seems ridiculous -- indeed almost meaningless. Aside from the obvious fact that we don't know what the "units" are units OF, we might well have gotten just as good results with 325, or 327 units, or indeed with 300 or 350 units. Since none of the units correspond to ANY particular aspect of the performance of the network, there is no particular justification for any one of them. Some might argue that the theory instantiated by the network is not meant to be read at this level of precision -- that it is not the number of units, specifically, that is being put forward for test, but only a network with a certain general sort of architecture and certain sorts of activation and learning rules. This seems simply too weak a claim to be of much scientific value. As Popper told us, scientists should put forward "bold conjectures" for test. The degree to which the hypothesis is subject to refutation by the test is the degree to which it is scientifically important. Even without accepting Popper's strong stand on the unique status of refutation in scientific work, this much remains clear: To back away from the details of one's theory -- to shield them from the possibility of refutation -- is to make one's theory scientifically less significant. Surely this is not a move connectionist researchers want to make in the long run. 12. It might be argued that the mapping of particular theoretical terms on to particular aspects of the behavior being modelled is unnecessary; it may just be an historical accident, primarily the result of our not being able to keep simultaneous control of thousands of theoretical terms until the advent of computers. Perhaps surprisingly, Carl Hempel seems to have presaged this possibility in his classic essay, Fundamentals of Concept Formation in Empirical Science: "A scientific theory might ... be likened to a complex spatial network: Its terms are represented by knots, while the threads connecting the latter correspond, in part, to the definitions and, in part, to the fundamental and derivative hypotheses included in the theory. The whole system floats, as it were, above the plane of observation and is anchored to it by rules of interpretation. These might be viewed as strings which are not part of the network but link to certain points of the latter with specific places in the plane of observation. By virtue of those interpretive connections, the network can function as a scientific theory: From certain observational data, we may ascend, via an interpretive string, to some point in the theoretical network, thence proceed, via definitions and hypotheses, to other points, from which another interpretive string permits a descent to the plane of observation." (Hempel, 1952, p. 36) 13. Now, it is by no means clear that Hempel had in mind here that there might be literally thousands of "knots in the network" between those few that are connected to the "plane of observation," but by the same token there is nothing in the passage that seems to definitely preclude the possibility either. 14. The real question seems to be about what one can really be said to have learned about the phenomenon of interest if one's model of that phenomenon contains far more terms that are not tied down to the "empirical plane," so to speak, than it does entities that are. Consider the following analogy: suppose that an historian wants to understand the events that lead up to political revolutions, so he tries to simulate several revolutions and a variety of other less successful political uprisings with a connectionist network. The input units encode data on, say, the state of the economy in the years prior to the uprising, the morale of the population, the kinds of political ideas popular at the time, and a host of other important socio- political variables. The output units encode various possible outcomes: revolution, uprising forcing significant political change, uprising diffused by superficial political concessions, uprising put down by force, etc. Among the input and output units, let us say that the historian places exactly 72 units which, he says, encode "a distributed representation of the socio-political situation of the time." His simulation runs beautifully. Indeed, let us say that because he has learned the latest techniques of recurrent networks, he is actually able to simulate events in the order in which they took place over several years either side of each uprising. 15. What has he learned about revolution? That there must have been (even approximately) 72 units involved? Certainly not. If the "hidden" units corresponded to something in particular -- say, to political leaders, or parties, or derivative socio-political variables -- that is, if the network had been SYMBOLIC, then perhaps he would have a case. Instead, he must simply repeat the mantra that they constitute "a distributed representation of the situation," and that the network is likely a close approximation to the situation because it plausibly simulates so many different variants of it. 16. It must be concluded that he has not learned very much about revolution at all. The simple fact of having a working "simulation" seems to mean little. It is only if one can interpret the INTERNAL ACTIVITY of the simulation that the simulation increases our knowledge; i.e., it is only then that the simulation is to be considered a scientific THEORY worthy of consideration. 17. Some might find this analogy invalid because of the widely recognized problems with studying history with the methods of science. My own opinion is that this is a non sequitur; but rather than arguing the point let us turn to a less controversial case. Assume for the moment that some aspiring amateur physicist, blithely unaware of the work of Galileo and Newton, gets the idea that the way to study the dynamics of balls rolling down inclined planes is to simulate their movements with a connectionist network. He sets up the net with inputs corresponding to variables such as the mass and volume of the ball, the length and angle of the plane, etc. Perhaps, not really knowing what he is after, he adds in some interesting variations such as ellipsoidal balls and curved surfaces, and includes the pertinent features of these in his encoding scheme. The activity of the output unit represents simply the time it takes the ball to complete its descent down the surface. He throws in a handful of hidden units, say 5, and runs the simulation. Eventually the network is able to predict closely how long it will take a certain ball to run down a certain surface, and it is able to generalize its knowledge to new instances on which it was not trained. If asked what the hidden units represent, the young physicist says, "the individual units represent nothing in particular; just a distributed representation of the physical situation as a whole." What has he learned? Not much, it would seem. Certainly not what was learned in the explanation of these kinds of phenomena in the theories of Galileo and Newton, in which the theoretical entities clearly REFER to relatively uncontroversial aspects of the world (e.g., distance, duration, size). 18. One way we cognitive scientists might try to avoid the fate of our hypothetical connectionist historian and physicist is to claim that connectionist units DO correspond to something closely related to the cognitive domain; viz., the neurons of the brain. Whether this is to be considered an analogy or an actual literal claim is often left vague by those who suggest it. Most connectionists seem wary of proclaiming too boldly that their networks model the actual activity of the brain. McClelland, Rumelhart, and Hinton (1986), for instance, say that connectionist models "seem much more closely tied to the physiology of the brain than other information-processing models" (p. 10), but then they retreat to saying that their "physiological plausibility and neural inspiration...are not the primary bases of their appeal to us" (p. 11). Smolensky (1988), after having examined a number of possible mappings, writes that "given the difficulty of precisely stating the neural counterpart of components of subsymbolic [i.e., connectionist] models, and given the very significant number of misses, even in the very general properties considered..., it seems advisable to keep the question open" (p. 9). Only with this caveat in place does he then go on to claim that "there seems no denying, however, that the subconceptual [i.e., connectionist] level is SIGNFICANTLY CLOSER [emphasis added] to the neural level than is the conceptual [i.e., symbolic] level" (p. 9). Precisely what metric he is using to measure the "closeness" of various theoretical approaches to the neural level of description is left unexplicated. 19. The general aversion to making very strong claims about the relation between connectionist models and brain is not without good reason. Crick and Asanuma (1986) describe five properties that the units of connectionist networks typically have that are rarely or never seen in neurons, and two further properties of neurons that are rarely found in the units of connectionist networks. Perhaps most important of these is the fact that the success of connectionist models seems to DEPEND upon the fact that any given unit can send excitatory impulses to some units and inhibitory impulses to others. No neuron in the mammalian brain is known to do this (though "dual-action" neurons have been found in the abdominal ganglion of Aplysia; see Levitan & Kaczmarek, 1991, pp. 196-197). Although it is certainly possible that dual-action neurons will be found in the human brain, the vast majority of cells do not seem to have this property, whereas the vast majority of units in connectionist networks typically do. Even as strong a promoter of connectionism as Paul Churchland (1990, p. 221) has recognized this as a major hurdle to be overcome if connectionist nets are to be taken seriously as models of brain activity. What is more, despite some obvious but possibly superficial similarities between the structure of connectionist units and the structure of neurons, there is currently little hard evidence that any SPECIFIC aspect of cognition is instantiated in the brain by neurons arranged in any SPECIFIED connectionist configuration. 20. It would accordingly appear that at present the only way of interpreting connectionist networks as serious candidates for theories of cognition, would be as literal models of the brain activity that underpins cognition. This means, if Crick and Asanuma are right in their critique, that connectionists should start restricting themselves to units, connections, and rules that use all and only principles that are known to be true of neurons. Other interpretations of connectionist networks may be possible in principle, but at this point none seem to have appeared on the intellectual horizon [4]. Without such an interpretation, connectionist modelers are left more or less in the position of out hypothetical connectionist historian. Even a simulation that is successful in terms of transforming certain inputs into the "right" outputs does not tell us much about the cognitive process it is simulating unless there is a plausible interpretation of its inner workings. All the researcher can claim is that the success of the simulation confirms that SOME connectionist architecture is involved, and perhaps something very general about the nature of that architecture (e.g., that it is self-organizing, recurrent, etc.). There is little or no confirmation of the specific features of the network because so much of it is OPTIONAL. 21. Now, it might be argued that this situation is no different from that of early atomic theory in physics. Visible bits of matter and their interactions with other bits of matter were explained by the postulation of not just thousands, but millions upon millions of theoretical entities of mostly unknown character -- viz., atoms. This, the argument would continue, is not so different from the situation in connectionism. After all, as Lakatos (1970) taught us, new research programs need a grace period in the beginning to get themselves established. Although I don't have a demonstrative argument against this line of thought, I think it has relatively little merit. We know pretty well what atoms are, and where we would find them, were we able to achieve the required optical resolution. Put very bluntly, if you simply look closer and closer and closer at a material object, you'll eventually see the atoms. Atoms are, at least in that sense, perfectly ordinary material objects themselves. Although they constitute an extension of our normal ontological categories, they do not REPLACE an old well-understood category with a new ill-understood one.[5] 22. By contrast, the units of connectionist networks (unless identified with neurons, or other bits of neural material) are quite different. They are not a REDUCTION of mental concepts, and as such give us no obvious path to follow to get from the "high level" of behavior and cognition to the "low level" of units and connections. That it is not a REDUCTIVE position is in fact often cited as a STRENGTH of connectionism but, if I am right, it is also the primary source of the ontological problems that have been discussed here. 23. To conclude, it is important to note that I am not arguing that connectionist networks must give way to symbolic networks because cognition is inherently symbolic (see, e.g., Fodor & Pylyshyn, 1988). That is an entirely independent question. What I am suggesting, however, is that the apparent success of connectionism in domains where symbolic models typically fail may be due as much to the huge number of additional "degrees of freedom" that connectionist networks are afforded by virtue of the blanket claim of distributed representation across large numbers of uninterpreted units, as it is to any inherent virtues that connectionism has over symbolism in explaining cognitive phenomena. FOOTNOTES [1] There are many ways of studying implicit memory. The experiment I describe here is, I believe, a "classic" procedure, but by no means the only one. [2] A Psycoloquy reviewer of this paper suggested that it is not the individual units that are theoretical entities, but only the units as a general type. He explicitly compared the situation to that of statistical dynamics, in which the phenomena are said to result from the actions of large, but unspecified, numbers of molecules of a general type. The difference is, of course, that we have lots of independent evidence of the existence of molecules. We know quite a lot about their properties. The same cannot be said of units in connectionist networks. Their existence is posited SOLELY for the purpose of making the networks behave the way we want them to. There is no independent evidence of their properties or their existence at all. [3] Notice that a version of the Sorites paradox threatens here. There must come a point where the subtraction of a unit from the network would lead to a decrement in its performance, but typically connectionist researchers work well above this level in order to optimize learning speed and generalization. [4] One Psycoloquy referee suggested that units might correspond to small neural circuits rather than individual neurons. This might be so, but the evidential burden is clearly on the person who makes this proposal to find some convincing empirical evidence for it. [5] There may be a temptation to attempt to carry this through to the quantum level, and claim that it does not carry through at that level because of the physical impossibility of seeing subatomic particles. First of all, relying on our intuitions about the quantum world to illuminate other scientific spheres is a very dangerous move because it is there more than anywhere that our intuitions seem to fail. Despite this, the move would fail in any case because the impossibility at issue is merely PHYSICAL, not LOGICAL. In a world in which light turned out to be continuous rather than particulate, the argument would carry through perfectly well. Put less technically, we know WHERE to see subatomic particles, we just don't know HOW to see them. The same cannot be said for units in connectionist networks. They simply don't seem to refer to ANYTHING in the system being studied at all. REFERENCES Atkinson, R. C. & Shiffrin, R. M. (1971) The control of short-term memory. Scientific American 225:82-90. Baddeley (1992) Working memory. Science 255:556-559. Berwick, R. C. (1985) The acquisition of syntactic knowledge, MIT Press. Churchland, P. M. (1990) Cognitive activity in artificial neural networks. In: Thinking: An invitation to cognitive science (Vol. 3), ed. D. N. Osherson & E. E. Smith, MIT Press. Fodor, J. A. & Pylyshyn, Z. W. (1988). Connectionism and cognitive architecture: A critical analysis. Cognition 28:3-71. Hempel, C. G. (1952) Fundamentals of concept formation in empirical science. University of Chicago Press. Jacoby, L. L. (1991) A process dissociation framework: Separating automatic from intentional uses of memory. Journal of Memory & Language 30: 513-541. Lakatos, I. (1970). Falsification and the methodology of scientific research programmes. In: Criticism and the growth of knowledge, ed. I. Lakatos & A. Musgrave (Eds.), Cambridge University Press. Levitan, I. B. & Kaczmarek, L. K.. (1991). The neuron: Cell and molecular biology. New York: Oxford University Press. McClelland, J. L., Rumelhart, D. E., & Hinton, G. E. (1986) The appeal of parallel distributed processing. In: Parallel distributed processing: Explorations in the microstructure of cognition (vol. 1), ed. Rumelhart, D. E. & McClelland, J. L., MIT Press. Roediger, H. L. III (1990) Implicit memory: Retention without remembering. American Psychologist 45:1043-1056. Roediger, H. L., III, & McDermott, K. B. (1993) Implicit memory in normal human subjects. In: Handbook of neuropsychology (Vol. 8, pp. 63-131), ed. F. Boller & J. Grafman, Elsevier. Schacter, D. L. (1987) Implicit memory: History and current status. Journal of Experimental Psychology: Learning, Memory, and Cognition 13:501-518. Schacter, D. L. (1992). Understanding implicit memory: A cognitive neuroscience approach. American Psychologist 47:559- 569. Smolensky, P. (1988). On the proper treatment of connectionism. Behavioral and Brain Sciences 11:1-73. Tulving, E. (1985) How many memory systems are there? American Psychologist 40:385-398. From db10 at cus.cam.ac.uk Wed Apr 15 10:36:39 1998 From: db10 at cus.cam.ac.uk (David Brown) Date: Wed, 15 Apr 1998 15:36:39 +0100 Subject: POSTDOC - NEURAL MODELLER - CAMBRIDGE, UK Message-ID: <3.0.5.32.19980415153639.008f8dc0@pop.cus.cam.ac.uk> POSTDOCTORAL FELLOWSHIP - NEURAL MODELLER Laboratory of Computational Neuroscience, The Babraham Institute, Cambridge, UK. MODELLING A DUAL-FUNCTION NEURAL NETWORK This EU funded project linking laboratories in Cambridge, Edinburgh, Montpellier and Rome will involve the construction of mathematical models of oxytocin neurones, assessing model properties by computer simulation and analytical techniques where possible, and participation in the planning and execution of experimental tests of the models, in conjunction with neurobiologists (at Montpellier and Edinburgh), e.g. by devising critical experiments to discriminate between hypotheses. Random synaptic input is an important element of the environment of oxytocin neurones, which have two modes of activity corresponding to different physiological functions: firing in synchronised, high frequency bursts during lactation and parturition; and in a continuous firing pattern, responding to the imposed osmotic stress by a graded increase in mean activity. Models will probably be required at various levels of complexity, spanning the range from leaky integrator models to biophysical models with dendritic and somatic compartments. The postdoctoral research worker appointed will work closely with Jianfeng Feng and David Brown in the Lab. of Computational Neuroscience at Babraham, and in collaboration with mathematicians in the Physics Department, Rome 'La Sapienza' University (whose main focus will be assembling the single neurone models developed at Cambridge into networks) and neuroendocrinologists in Montpellier and Edinburgh. This is an exciting and novel project. The network's capacity for two distinct modes of action, both physiologically important, in response to different inputs will probably require the development of new models. Because of the low-dimensionality of system outputs (either tonic or pulsatile hormone release), sufficient good experimental data (e.g. simultaneous electrophysiological recordings, hormone release etc) can be collected within the project for thorough experimental calibration and testing of models. The biologists involved have between them an unrivalled experience and knowledge of the oxytocin system, and are pioneering cutting-edge experimental techniques for its study. The person appointed should have a PhD or equivalent research experience in biological, preferably neuronal or physiological modelling, with knowledge of analytical and simulation based techniques for assessing the behaviour of neuronal models, and relating the models to experimental data. A good first degree in a mathematically based subject and experience of biological computing will probably be required. The project will involve some travel between the four sites, so as to facilitate regular contact with the mathematicians and biologists involved. Salary in approx. range ?16,000-?26,000 per annum depending on qualifications and experience. The project will start in mid-1998, and continue in the first instance for 2 years. Further information from David Brown (+44 (0)1223 832312, Fax +44 (0)1223 837912, email db10 at cus.cam.ac.uk). Applications in the form of a CV and names and addresses of three referees to David Brown, Laboratory of Computational Neuroscience, Babraham Institute, Cambridge CB2 4AT as soon as possible and at the latest by 30th May 1998. From franz at homer.njit.edu Fri Apr 17 10:55:49 1998 From: franz at homer.njit.edu (Franz Kurfess) Date: Fri, 17 Apr 1998 10:55:49 -0400 Subject: CfP Special Issue "Neural Networks and Structured Knowledge" In-Reply-To: <40984038@toto.iv> Message-ID: <199804171455.KAA11297@vector.njit.edu> We received a number of requests to extend the deadline for the Special Issue "Neural Networks and Structured Knowledge" in "Applied Intelligence", and decided to revise the schedule, in effect pushing all dates back by two months. Here is the new schedule: Revised Schedule Paper submission deadline: July 1, 1998 Review decision by: August 31, 1998 Final manuscript due: September 31, 1998 Tentative publication date: January 1999 I'm appending the Call for Contributions for your reference. Best regards, Franz Kurfess, Guest Editor Special Issue "Neural Networks and Structured Knowledge" in Applied Intelligence: The International Journal of Artificial Intelligence, Neural Networks, and Complex Problem-Solving Techniques Call for Contributions The submission of papers is invited for a special issue on "Neural Networks and Structured Knowledge" of the Applied Intelligence Journal. Issue Theme The representation and processing of knowledge in computers traditionally has been concentrated on symbol-oriented approaches, where knowledge items are associated with symbols. These symbols are grouped into structures, reflecting the important relationships between the knowledge items. Processing of knowledge then consists of manipulation of the symbolic structures, and the result of the manipulations can be interpreted by the user. Whereas this approach has seen some remarkable successes, there are also domains and problems where it does not seem adequate. Some of the problems are computational complexity, rigidity of the representation, the difficulty of reconciling the artificial model with the real world, the integration of learning into the model, and the treatment of incomplete or uncertain knowledge. Neural networks, on the other hand, have advantages that make them good candidates for overcoming some of the above problems. Whereas approaches to use neural networks for the representation and processing of structured knowledge have been around for quite some time, especially in the area of connectionism, they frequently suffer from problems with expressiveness, knowledge acquisition, adaptivity and learning, or human interpretation. In the last years much progress has been made in the theoretical understanding and the construction of neural systems capable of representing and processing structured knowledge in an adequate way, while maintaining essential capabilities of neural networks such as learning, tolerance of noise, treatment of inconsistencies, and parallel operation. The theme of this special issue comprises * the investigation of the underlying theorecical foundations, * the implementation and evaluation of methods for representation and processing of structured knowledge with neural networks, and * applications of such approaches in various domains. Topics of Interest The list below gives some examples of intended topics. * Concepts and Methods: o extraction, injection and refinement of structured knowledge from, into and by neural networks o inductive discovery/formation of structured knowledge o combining symbolic machine learning techniques with neural lerning paradigms to improve performance o classification, recognition, prediction, matching and manipulation of structured information o neural methods that use or discover structural similarities o neural models to infer hierachical categories o structuring of network architectures: methods for introducing coarse-grained structure into networks, unsupervised learning of internal modularity * Application Areas: o medical and technical diagnosis: discovery and manipulation of structured dependencies, constraints, explanations o molecular biology and chemistry: prediction of molecular structure unfolding, classification of chemical structures, DNA analysis o automated reasoning: robust matching, manipulation of logical terms, proof plans, search space reduction o software engineering: quality testing, modularisation of software o geometrical and spatial reasoning: robotics, structured representation of objects in space, figure animation, layouting of objects o other applications that use, generate or manipulate structures with neural methods: strucures in music composition, legal reasoning, architectures, technical configuration, ... The central theme of this issue will be the treatment of structured information using neural networks, independent of the particular network type or processing paradigm. Thus the theme is orthogonal to the question of connectionist/symbolic integration, and is not intended as a continuation of the more philosphically oriented discussion of symbolic vs. subsymbolic representation and processing. Submission Process Prospective authors should send an electronic mail message indicating their intent to submit a paper to the guest editor of the special issue, Franz J. Kurfess (kurfess at cis.njit.edu). This message should contain a preliminary abstract and three to five keywords. Six hard copies of the final manuscript should be sent to the guest editor (not to the Applied Intelligence Editorial office): Prof. Franz J. Kurfess New Jersey Institute of Technology Phone: (973) 596 5767 Department of Computer and Information Science Fax: (973) 596 5777 University Heights Email: kurfess at cis.njit.edu Newark, NJ 07102-1982 WWW: http://www.cis.njit.edu/~franz To speed up the reviewing process, authors should also send a PostScript version of the paper via email to the guest editor. Prospective authors can find further information about the journal on the home page http://kapis.www.wkap.nl/journalhome.htm/0924-669X Schedule Paper submission deadline: July 1, 1998 Review decision by: August 31, 1998 Final manuscript due: September 31, 1998 Tentative publication date: January 1999 From mieko at hip.atr.co.jp Fri Apr 17 07:11:26 1998 From: mieko at hip.atr.co.jp (Mieko Namba) Date: Fri, 17 Apr 1998 20:11:26 +0900 Subject: CALL FOR PAPERS [Neural Networks 1999 Special Issue] Message-ID: <199804171111.UAA04869@mailhost.hip.atr.co.jp> Dear members, We are glad to inform you that the Japanese Neural Networks Society will edit the NEURAL NETWORKS 1999 Special Issue as below. NEURAL NETWORKS is an official international compilation of the Journal of the International Neural Networks Society, the European Neural Networks Society and the Japanese Neural Networks Society. We are looking forward to receiving your contributions. Mitsuo Kawato Co-Editor-in-Chief Neural Networks (ATR Human Information Proc. Res. Labs.) ****************************************************************** CALL FOR PAPERS ****************************************************************** Neural Networks 1999 Special Issue "Organisation of Computation in Brain-like Systems" ****************************************************************** Co-Editors: Professor Gen Matsumoto, BSI, RIKEN, Japan Professor Edgar Koerner, HONDA R&D, Europe Dr. Mitsuo Kawato, ATR Human Information Processing Res. Labs., Japan Submission: Deadline for submission: December 1st, 1998 Notification of acceptance: March 1st, 1999 Format: as for normal papers in the journal (APA format) and no longer than 10,000 words Address for Papers: Dr. Mitsuo Kawato ATR Human Information Processing Research Laboratories 2-2 Hikaridai, Seika-cho Soraku-gun, Kyoto 619-0288, Japan. ****************************************************************** In the recent years, neuroscience has made a big leap forward regarding both investigation methodology and insights in local mechanisms of processing of sensory information in the brain. The fact that we still do not know much better than before what happens in the brain when one recognises a familiar person, or moves around navigating seemingly effortless through a busy street, points to the fact that our models still do not describe essential aspects of how the brain organises computation. The investigation of the behaviour of fairly homogeneous ANS (artificial neural systems) composed of simple elementary nodes fostered the awareness that architecture matters: Algorithms implemented by the respective neural system are expressed by its architecture. Consequently, the focus is shifting to better understanding of the architecture of the brain and of its subsystems, since the structure of those highly modularised systems represents the way the brain organises computation. Approaching the algorithms expressed by those architectures may offer us the capability to not only understand the representation of knowledge in a neural system made under well defined constraints, but to understand the control that forces the neural system to make representations of behaviourally relevant knowledge by generating dynamic constraints. This special issue will bring together invited papers and contributed articles that illustrate the shifting emphasis in neural systems modelling to more neuroarchitecture-motivated systems that include this type of control architectures. Local and global control algorithms for organisation of computation in brain-like systems cover a wide field of topics. Abduction of control principles inherent in the architectures that mediate interaction within the cortex, between cortex -thalamus, cortex-hippocampus and other parts of the limbic system is one of the targets. Of particular importance are the rapid access to stored knowledge and the management of conflicts in response to sensory input, the coding and representation in a basically asynchronous mode of processing, the decomposition of problems into a reasonable number of simpler sub-problems, and the control of learning -- including the control which specifies what should be learned, and how to integrate the new knowledge into the relational architecture of the already acquired knowledge representation. Another target of that approach is the attempt to understand how these controls and the respective architectures emerged in the process of self-organisation in the phylogenetic and ontogenetic development. Setting the cognitive behaviour of neural systems in the focus of investigation is a prerequisite for the described approach that will promote both creating computational hypotheses for neurobiology and implementing robust and flexible computation in ANS. ****************************************************************** end. ========================================================= Mieko Namba Secretary to Dr. Mitsuo Kawato Editorial Administrator of NEURAL NETWORKS ATR Human Information Processing Research Laboratories 2-2 Hikaridai, Seika-cho, Soraku-gun, Kyoto 619-0288, Japan TEL +81-774-95-1058 FAX +81-774-95-1008 E-MAIL mieko at hip.atr.co.jp ========================================================= From pmitra at bell-labs.com Mon Apr 20 22:58:09 1998 From: pmitra at bell-labs.com (Partha Mitra) Date: Mon, 20 Apr 1998 22:58:09 -0400 Subject: Please distribute Message-ID: <353C0B41.5C03@bell-labs.com> _______________________ Analysis of Neural Data _______________________ Modern methods and open issues in the analysis and interpretation of multi-variate time-series and imaging data in the neurosciences ___________________________________________________ >> 16 August - 29 August 1998 >> Marine Biological Laboratories - Woods Hole, MA ___________________________________________________ A working group of scientists committed to quantitative approaches to problems in neuroscience will focus their efforts on experimental and theoretical issues related to the analysis of large, single- and multi-channel data sets. The motivation for the work group is based on issues that arise in two complimentary areas critical to an understanding of brain function. The first involves advanced signal processing methods, particularly those appropriate for emerging multi-site recording techniques and noninvasive imaging techniques. The second involves the development of a calculus to study the dynamical behavior of nervous systems and the computations they perform. A distinguishing feature of the work group will be the close collaboration between experimentalists and theorists, particularly with regard to the analysis of data and the planning of experiments. The work group will have a limited number of research lectures, supplemented by tutorials on relevant computational, experimental, and mathematical techniques. This work group is a means to critically evaluate techniques for the processing of multi-channel data, of which imaging forms an important category. Such techniques are of fundamental importance for basic research and medical diagnostics. We will establish a repository of these techniques, along with benchmarks, to insure the rapidly dissemination of modern analytical techniques throughout the neuroscience community. The work group will convene on a yearly basis. For 1997, we propose to focus on topics that fall under the rubric of multivariate time-series. * Analysis of point processes, e.g., spike trains. Measures of correlation and variability, and their interpretation. * Analysis of continuous processes, e.g., field potential, optical imaging, fMRI, and MEG, and the recording of behavioral output, e.g., vocalizations. * Problems that involve both point and continuous processes, e.g., the linear and nonlinear functional relations between spike trains and sensory input and motor output. Participants: Twenty five participants, both experimentalists and theorists. Experimentalists are specifically encouraged to bring data records to the work group; appropriate computational facilities will be provided. The work group will further take advantage of interested investigators and course faculty concurrently present at the MBL. We encourage graduate students and postdoctoral fellows as well as senior researchers to apply. Participant Fee: $200. Support: National Institutes of Health - NIMH, NIA, NIAAA, NICHD/NCRR, NIDCD, NIDA, and NINDS. Organizers: David Kleinfeld (UCSD) and Partha P. Mitra (Caltech and Bell Laboratories). Website: http://www-physics.ucsd.edu/research/neurodata Application: Send a copy of your curriculum vita, together with a cover letter that contains a brief (ca. 200 word) paragraph on why you wish to attend the work group and a justified request for any financial aid, to: Ms. Jean B. Ainge Bell Laboratories, Lucent Technologies 700 Mountain Avenue 1D-427 Murray Hill, NJ 07974 908-582-4702 (fax) or The MBL is an EEO AAI. Graduate students and postdoctoral fellows are encouraged to include a brief letter of support from their research advisor. Financial assistance: Assistance for travel, accommodations, and board is available based on need. Applications must be received by 18 May 1998. Participants will be notified by 25 May Links to Archives for Neurosciences can be found at: http://www-physics.ucsd.edu/research/neurodata/NSarchive2.html From Wulfram.Gerstner at epfl.ch Wed Apr 22 09:49:39 1998 From: Wulfram.Gerstner at epfl.ch (Wulfram Gerstner) Date: Wed, 22 Apr 1998 15:49:39 +0200 (MET DST) Subject: preprints_on_spiking_neurons Message-ID: <199804221349.PAA06196@mantrasun8.epfl.ch> Three review papers on Neural Networks with Spiking Neurons can be retrieved from the following web page. http://diwww.epfl.ch/lami/team/gerstner/wg_pub.html --------------------------------------------------- I. SPIKING NEURONS. (W. Gerstner) This tutorial paper gives a review of several models of spiking neurons (integrate-and-fire; Hodgkin-Huxley; spike response model). (55 pages) II. POPULATIONS OF SPIKING NEURONS. (W. Gerstner) The second paper develops the mathematical framework to describe populations of spiking neurons. Specific topics are (i) the rapid response of populations of spiking neurons to changes in the input. (ii) exact stability conditions for perfectly synchronized locked solutions as well as (iii) the stability of incoherent firing activity in the presence of noise and transmission delays. (38 pages) III. HEBBIAN LEARNING OF PULSE TIMING IN THE BARN OWL AUDITORY SYSTEM. (W. Gerstner, R. Kempter, J.L. van Hemmen and H. Wagner) A correlation based learning rule based on spike timing is discussed and applied to the problem of coincidence detection and sound localization in the barn owl auditory system. Specific topics are (i) the relation of spike-based and rate-based learning (ii) delay line selection by learning (iii) phase locking in the auditory system. (26 pages) ----------------------------------------------------- The three papers are preprints of book chapters. The book entitled Pulsed Neural Nets edited by W. Maas and C. Bishop will appear in October/November 98 (MIT Press). From dror at coglit.soton.ac.uk Wed Apr 22 18:28:52 1998 From: dror at coglit.soton.ac.uk (Itiel Dror) Date: Wed, 22 Apr 1998 23:28:52 +0100 (BST) Subject: Position at Southampton University Message-ID: I would appreciate it very much if you could please post the following job announcement. Thank you. Itiel #======================================================================# | Itiel E. Dror, Ph.D. http://www.cogsci.soton.ac.uk/~dror/ | | Department of Psychology dror at coglab.psy.soton.ac.uk | | University of Southampton Office 44 (0)1703 594519 | | Highfield, Southampton Lab. 44 (0)1703 594518 | | England SO17 1BJ Fax. 44 (0)1703 594597 | #======================================================================# ******************************************************************************* UNIVERSITY OF SOUTHAMPTON DEPARTMENT OF PSYCHOLOGY FULL PROFESSOR IN COGNITIVE PSYCHOLOGY Applications are invited for a Full Professor in Cognitive Psychology in the Department of Psychology tenable as soon as possible. In a recent review of our research activity the University identified the further strengthening of the Cognitive Psychology base within the Department as a strategic priority. As a result the University has provided additional resources to support developments in this field. The established Professorship which falls vacant in September this year has been specifically designated as a post in Cognitive Psychology. In addition, two new junior position have been created to support the appointment of the new Full Professor. The junior positions will be advertised only after the appointment of the Full Professor in order to promote complementarity amongst our Cognitive Psychology grouping. Substantial funds will be set aside for the purchase of equipment to support the research of the successful applicant. The Department has recently moved into refurbished premises in the Shackleton Building which provide flexible research space with extensive laboratory facilities for experimental work. The new Professor in Cognitive Psychology will be given lighter teaching and administrative duties during their first year than would normally be the case. These new posts in Cognitive Psychology are part of a larger package of new HEFCE funded appointments. There will also be a new Full Professor and a new junior position in Social Psychology, two senior positions (Health and Developmental Psychology) and a junior position in Human Learning and Behaviour Analysis. Each post has been targeted to play a specific role in relation to one of the new Research Groups identified during the review mentioned above. Technical and clerical support within the Department is also being overhauled. These changes will lead to a further period of growth and development within the Department and further strengthen what is already a lively and effective research team. At the same time they will increase the scope of teaching at both the undergraduate and postgraduate level. These developments also provide a clear indication of the University of Southampton's support for Psychology's commitment to become a major center of excellence in research and teaching. The new posts in Cognitive Psychology will strengthen what is already a productive group of young cognitive scientists working within the Department. The successful candidate will be expected to provide leadership in this area and so play a major role in the development of the recently established Cognitive Psychology Research Group. An essential requirement is that he or she will have the capacity to facilitate the research of others. Collaboration with colleagues within and outside the Cognitive Psychology grouping will be encouraged. The appointee's research could be in any area of Cognitive Psychology. However, he or she will have an outstanding track-record of achievement in theoretically driven empirical research of a fundamental nature evidenced by publication in the leading international refereed journals. The Department would welcome informal enquiries and visits. Potential applicants should contact the Head of Department, Professor Edmund Sonuga-Barke, on (01703) 594606 or esb at psy.soton.ac.uk. or to Dr Itiel Dror (dror at coglab.psy.soton.ac.uk) or Dr Sarah Stevenage (svs1 at soton.ac.uk) From dld at cs.monash.edu.au Thu Apr 23 07:28:11 1998 From: dld at cs.monash.edu.au (David L Dowe) Date: Thu, 23 Apr 1998 21:28:11 +1000 Subject: CFPs: Information theory in biology, Jan 99, Hawaii Message-ID: <199804231128.VAA11139@dec11.cs.monash.edu.au> Dear All, Apologies for cross-postings. In short, if you're interested in MML or MDL or Akaike's Information Criterion or information theory and you're interested in biology, and you'd like to go to Hawaii in January 1999, and you can get a paper ready by July 1998, then store this mail away and bookmark the site http://www.cs.monash.edu.au/~dld/PSB99/PSB99.Info.CFPs.html and read on. =================================================================== This is the Call For Papers for the 4th Pacific Symposium on BioComputing (PSB99, 1999) conference track on "Information-theoretic approaches to biology". PSB-99 will be held from 4-9 January, 1999, in Mauni Lani on the Big Island of Hawaii. Track Organisers: David L. Dowe (dld at cs.monash.edu.au) and Klaus Prank. WWW site: http://www.cs.monash.edu.au/~dld/PSB99/PSB99.Info.CFPs.html . Specific technical area to be covered by this track: Approaches to biological problems using notions of information or complexity, including methods such as Algorithmic Probability, Minimum Message Length and Minimum Description Length. Two possible applications are (e.g.) protein folding and biological information processing. Kolmogorov (1965) and Chaitin (1966) studied the notions of complexity and randomness, with Solomonoff (1964), Wallace (1968) and Rissanen (1978) applying these to problems of statistical and inferential learning (and ``data mining'') and to prediction. The methods of Solomonoff, Wallace and Rissanen have respectively come to be known as Algorithmic Probability (ALP), Minimum Message Length (MML) and Minimum Description Length (MDL). All of these methods relate to information theory, and can also be thought of in terms of Shannon's information theory, and can also be thought of in terms of Boltzmann's thermo-dynamic entropy. An MDL/MML perspective has been suggested by a number of authors in the context of approximating unknown functions with some parametric approximation scheme (such as a neural network). The designated measure to optimize under this scheme combines an estimate of the cost of misfit with an estimate of the cost of describing the parametric approximation (Akaike 1973, Rissanen 1978, Barron and Barron 1988, Wallace and Boulton, 1968). This track invites all original papers of a biological nature which use notions of information and/or information-theoretic complexity, with no strong preference as to what specific nature. Such work has been done in problems of, e.g., protein folding and DNA string alignment. As we shortly describe in some detail, such work has also been done in the analysis of temporal dynamics in biology such as neural spike trains and endocrine (hormonal) time series analysis using the MDL principle in the context of neural networks and context-free grammar complexity. To elaborate on one of the relevant topics above, in the last three years or so, there has been a major focus on the aspect of timing in biological information processing ranging from fields such as neuroscience to endocrinology. The latest work on information processing at the single-cell level using computational as well as experimental approaches reveals previously unimagined complexity and dynamism. Timing in biological information processing on the single-cell level as well as on the systems level has been studied by signal-processing and information-theoretic approaches in particular in the field of neuroscience (see for an overview: Rieke et al. 1996). Using such approaches to the understanding of temporal complexity in biological information transfer, the maximum information rates and the precision of spike timing to the understanding of temporal complexity in biological information transfer, the maximum information rates and the precision of spike timing could be revealed by computational methods (Mainen and Sejnowski, 1995; Gabbiani and Koch 1996; Gabbiani et al., 1996). The examples given above are examples of some possible biological application domains. We invite and solicit papers in all areas of (computational) biology which make use of ALP, MDL, MML and/or other notions of information and information-theoretic complexity. In problems of prediction, as well as using "yes"/"no" predictions, we would encourage the authors to consider also using probabilistic prediction, where the score assigned to a probabilistic prediction is given according to the negative logarithm of the stated probability of the event. Further comments re PSB-99 : ---------------------------- PSB99 will publish accepted full papers in an archival Proceedings. All contributed papers will be rigorously peer-reviewed by at least three referees. Each accepted full paper will be allocated up to 12 pages in the conference Proceedings. The best papers will be selected for a 30-minute oral presentation to the full assembled conference. Accepted poster abstracts will be distributed at the conference separately from the archival Proceedings. To be eligible for proceedings publication, each full paper must be accompanied by a cover letter stating that it contains original unpublished results not currently under consideration elsewhere. See http://www.cgl.ucsf.edu/psb/cfp.html for more information. IMPORTANT DATES: Full paper submissions due: July 13, 1998 Poster abstracts due: August 22, 1998 Notification of paper acceptance: September 22, 1998 Camera-ready copy due: October 1, 1998 Conference: January 4 - 9, 1999 More information about the "Information-theoretic approaches to biology" track, including a sample list of relevant papers is available on the WWW at http://www.cs.monash.edu.au/~dld/PSB99/PSB99.Info.CFPs.html . More information about PSB99 is available from http://www.cgl.ucsf.edu/psb/cfp.html For further information, e-mail Dr. David Dowe, dld at cs.monash.edu.au or e-mail Dr. Klaus Prank, ndxdpran at rrzn-serv.de . This page was put together by Dr. David Dowe, School of Computer Science and Softw. Eng., Monash University, Clayton, Vic. 3168, Australia e-mail: dld at cs.monash.edu.au Fax: +61 3 9905-5146 http://www.csse.monash.edu.au/~dld/ and Dr. Klaus Prank, Abteilung Klinische Endokrinologie Medizinische Hochschule Hannover Carl-Neuberg-Str. 1 D-30623 Hannover Germany e-mail: ndxdpran at rrzn-serv.de Tel.: +49 (511) 532-3827 Fax.: +49 (511) 532-3825 http://sun1.rrzn-user.uni-hannover.de/~ndxdpran/ From Paul.Keller at pnl.gov Thu Apr 23 13:26:12 1998 From: Paul.Keller at pnl.gov (Keller, Paul E) Date: Thu, 23 Apr 1998 10:26:12 -0700 Subject: Job Announcement: Cognitive Systems Engineer at Battelle Message-ID: <7A8CF1DC6A9DD0118EA400A024BF29DA0229BB89@pnlmse2.pnl.gov> Cognitive Systems Engineer Battelle, a leading provider of technology solutions, has immediate need for a research engineer to join their cognitive systems initiative in their Columbus, Ohio, USA facility. The new position will provide technical support to a multi-year corporate project applying adaptive/cognitive information technology to applications in emerging technology areas. The position requires a B.S./M.S. in Computer and Information Science, Electrical Engineering, or related field with a specialization or experience in artificial neural networks, fuzzy logic, evolutionary computing/genetic algorithms, and statistical methods. In addition, the position requires intimate knowledge of Matlab, C/C++ language and object-oriented programming. Oral, written, and interpersonal communications skills are essential to this highly interactive position. The applicant selected will be subject to a security investigation and must meet eligibility requirements for access to classified information. Battelle offers competitive salaries, comprehensive benefits, and opportunities for professional development. Qualified candidates are invited to send their resumes to Battelle, Dept. J-92, 505 King Avenue, Columbus, OH 43201-2693 or e-mail them to priddy at battelle.org. Battelle is an Equal Opportunity/Affirmative Action Employer M/F/D/V. To find out more information about Battelle, try http://www.battelle.org. *** PLEASE RESPOND TO THE ADDRESS GIVEN ABOVE OR E-MAIL KEVIN PRIDDY AT priddy at battelle.org *** From n at predict.com Fri Apr 24 12:10:53 1998 From: n at predict.com (Norman Packard) Date: Fri, 24 Apr 1998 10:10:53 -0600 Subject: Job Announcement: research/software at Prediction Company Message-ID: <199804241610.KAA07707@seldon> PREDICTION COMPANY RESEARCH/SOFTWARE POSITION IN NONLINEAR MODELING OF FINANCIAL MARKETS April, 1998 Prediction Company is a small firm based in Santa Fe, NM, utilizing nonlinear forecasting technologies for prediction and computerized trading of financial instruments. We are seeking someone that can play a strong role in both research and research software. The basic task is to build models based on historical data to trade in financial markets. Responsibilities include application of existing technology, research and development of new technology, and participation in the design and building of an advanced software platform for prediction, trading, and risk control. The successful applicant for this job will have a Ph.D. in statistics, econonimcs, computer science, physics, mathematics, or a related field. Experience using time series modeling, machine learning, and statistical and numerical anlysis is highly valuable. Software experience is essential, particularly desirable in C++, S+ or related languages. Familiarity with finance is highly desirable. Experience with real data is also highly desirable -- the nastier the better. We are willing to consider a range of experience levels, including recent Ph.D.'s. The applicant should be willing to work in close collaboration with other researchers and software developers, and should be willing to take on what is unquestionably the most challenging but lucrative forecasting problem in existence. Prediction Company offers a relaxed and informal work environment. We are located in an historic three story building in the Guadalupe commercial district. Our offices include a full kitchen and roof deck. We are within easy walking distance of many cafes, restaurants and the historical central plaza of Santa Fe. For further information check out our web page at www.predict.com. Applicants should email resumes to Laura Barela at laura at predict.com (postscript or ascii) or send by US mail to: Prediction Company Attn: Recruiting 236 Montezuma Avenue Santa Fe, NM 87501 From niall at zeus.csis.ul.ie Mon Apr 27 06:28:39 1998 From: niall at zeus.csis.ul.ie (Niall Griffith) Date: Mon, 27 Apr 1998 11:28:39 +0100 Subject: IEE Colloqiuim - Neural Nets and MultiMedia Message-ID: <9804271028.AA17535@zeus.csis.ul.ie> I am sorry if you receive this twice - according to a message that I have received it has not been delivered to you so I am trying again Niall Griffith My message was..... -------------------------------------------------------------- Please pass this on to anyone or any group you think may be interested. ============================================================== IEE Colloquium on "Neural Networks in Multimedia Interactive Systems" Thursday 22 October 1998, Savoy Place, London. Call for Papers - --------------- The IEE are holding a colloquium at Savoy Place on the use of neural network models in multimedia systems. This is a developing field of importance to both Multimedia applications developers who want to develop more responsive and adaptive systems as well as to neural network researchers. The aim of the colloquium is to present a range of current neural network applications in the area of interactive multimedia. The aim is cover a range of topics including learning, intelligent agents within multimedia systems, data mining, image processing and intelligent application interfaces. Invited Speakers: - ----------------- Bruce Blumberg, MIT Media Lab. Jim Austin, York. Russell Beale, Birmingham. Call For Papers - --------------- Submissions are invited in any (but not exclusively) of the following areas: Adaptive and plastic behaviour in multi-media systems Concept and behaviour learning and acquisition Browsing mechanisms Preference and strategy identification and learning Data mining Image processing in multimedia systems Cross modal and media representations and processes Intelligent agents Interested parties are invited to submit a two page (maximum) abstract of their proposed talk to either Dr. Niall Griffith, Department of Computer Science and Information Science, University of Limerick, Limerick, Ireland. email: niall.griffith at ul.ie Telephone: +353 61 202785 Fax: +353 61 330876 or Professor Nigel M Allinson Dept. of Elec. Eng. & Electronics UMIST PO Box 88 Manchester, M60 1QD, UK Voice: (+44) (0) 161-200-4641 Fax: (+44) (0) 161-200-4781/4 Internet: allinson at umist.ac.uk Timetable: - ---------- 29th April: Deadline for talk submissions 15th June: Authors notified. 24th November: Colloquium at IEE, Savoy Place, London ===================================================== - --RAB11527.891386128/oz.memphis.edu-- ------- End of forwarded message ------- From harnad at coglit.soton.ac.uk Mon Apr 27 09:24:08 1998 From: harnad at coglit.soton.ac.uk (Stevan Harnad) Date: Mon, 27 Apr 1998 14:24:08 +0100 (BST) Subject: Contribute to Ongoing Psyc Commentary on Green Message-ID: There is lively Commentary on Green's target article appearing in Psycoloquy, a refereed electronic journal sponsored by the American psychological Association. Further Commentary is invited. (All submissions are refereed.) URLs: US: http://www.princeton.edu/~harnad/psyc.html UK: http://www.cogsci.soton.ac.uk/psyc Address for submitting commentaries: psyc at pucc.princeton.edu Instructions at bottom of this message, preceded by latest commentary. Green, CD. Are Connectionist Models Theories of Cognition? PSYCOLOQUY 9(04) Tuesday 14 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.04.connectionist-explanation.1.green Orbach, J. Do Wires Model Neurons? PSYCOLOQUY 9(05) Wednesday 15 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.05.connectionist-explanation.2.orbach O'Brien, GJ. The Role of Implementation in Connectionist Explanation. PSYCOLOQUY 9(06) Sunday 19 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.06.connectionist-explanation.3.obrien Green, CD. Lashley's Lesson Is Not Germane. Reply to Orbach PSYCOLOQUY 9(07) Wednesday 22 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.07.connectionist-explanation.4.green Green, CD. Problems with the Implementation Argument. Reply to O'Brien PSYCOLOQUY 9(08) Saturday 25 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.08.connectionist-explanation.5.green Young, ME. Are Hypothetical Constructs Preferred Over Intervening Variables? PSYCOLOQUY 9(09) Monday 27 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.09.connectionist-explanation.6.young Grainger, J. & Jacobs, AM. Localist Connectionism Fits the Bill PSYCOLOQUY 9(09) Monday 27 April 1998 ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.10.connectionist-explanation.7.grainger ---------- psycoloquy.98.9.10.connectionist-explanation.7.grainger Mon 27 Apr 1998 ISSN 1055-0143 (6 paragraphs, 8 references, 153 lines) PSYCOLOQUY is sponsored by the American Psychological Association (APA) Copyright 1998 Jonathan Grainger LOCALIST CONNECTIONISM FITS THE BILL Commentary on Green on Connectionist-Explanation Jonathan Grainger Centre de Recherche en Psychologie Cognitive, CNRS Universite de Provence Aix-en-Provence France grainger at newsup.univ-mrs.fr Arthur M. Jacobs Dept. of Psychology Philips University of Marburg, Marbug, Germany jacobsa at mailer.uni-marburg.de ABSTRACT: Green (1998) restates a now standard critique of connectionist models: they have poor explanatory value as a result of their opaque functioning. However, this problem only arises in connectionist models that use distributed hidden unit representations, and is NOT a feature of localist connectionism. Indeed, Green's critique reads as an appeal for the development of localist connectionist models as an excellent starting point for building a unified theory of human cognition. 1. First, if we agree that theory development in psychological science is ready for the shift from prequantitative verbal-boxological modeling toward more formal modeling efforts, then the kinds of questions we should be asking are: What kind of quantitative modeling is appropriate? How should we evaluate its appropriateness? In other words, the verbal theories of human memory discussed by Green (1998) are not a serious alternative to whatever connectionism might offer. They are at best a starting point for developing more formal accounts of human memory. We have recently argued that localist connectionism provides a promising framework for such an endeavor (Grainger & Jacobs, 1998). 2. Green (1998), as well as many other critics of connectionism, appears to use the term connectionism as synonymous with trainable networks with hidden units (often called PDP models, and typically trained with backpropagation, Rumelhart, Hinton, & Williams, 1986). Many connectionist models do not include hidden units. Some of these are trainable (with Hebbian learning, for example), and some are hardwired (e.g., McClelland & Rumelhart's, 1981, interactive activation model). We refer to any connectionist model in which all processing units can be unambiguously assigned a meaningful interpretation as "localist connectionist." Note that, as in all connectionist models, all processing units in localist connectionist models are identical; it is only their position in the network that guarantees their unique interpretation. The modeler can artificially label each of these units in order to facilitate interpretation of network activity. 3. Grainger and Jacobs (1998) analyzed the advantages of adopting a localist connectionist approach as opposed to the currently more popular PDP approach. Here we will discuss only those points relevant to the issues raised by Green (1998). Green identifies the close connection between theoretical and observable entities as a critical feature of traditional scientific theories. One must be able to link transparently the theoretical entities of the theory to the observable entities in the target world in order to achieve explanatory adequacy. Without examining the extent to which this is fails to be a feature of PDP models, it should be clear from the above discussion that localist connectionist models do provide this transparent link. Units in localist connectionist models do refer to relatively uncontroversial aspects of the target world. They represent the categories (such as letters and words) that the brain has learned from repeated exposure to the environment. 4. As noted by Jacobs, Rey, Ziegler, and Grainger (1998), transparency will always tend to diminish as models become more complex. Jacobs et al. conclude, however, that algorithmic models of the localist connectionist variety may offer the best trade-off between clarity/transparency and formality/precision. It is the increased level of precision that allows localist connectionist models to achieve greater descriptive adequacy (Jacobs & Grainger, 1994) without sacrificing explanatory adequacy. 5. Apart from greater explanatory and descriptive adequacy, localist connectionist models offer a simple means of quantifying pre-existing verbal-boxological models that have already stood the test of extensive empirical research. Referring to this point, Page and Norris (1998) speak of a symbiosis between verbal theorizing and quantitative modeling. Furthermore, the principle of nested modeling has been readily applied with localist connectionist models. Adopting this approach facilitates the process of model-to-model comparison. Models differing by a single feature (e.g., interactivity, Jacobs & Grainger, 1992), can be compared, and different variants of the model can compete in strong inference studies (e.g., Dijkstra & van Heuven, 1998). 6. Finally, localist connectionist models, using the same simple processing units and activation functions, provide a unified explanation for phenomena observed in the different subdomains of human cognition. The general principles that govern processing in all localist models (e.g., similarity based parallel activation, lateral inhibition) can also be isolated and analyzed in an easily interpretable manner (see e.g., Grainger & Jacobs, in press). We therefore conclude that localist connectionism provides an excellent starting point for the development of a unified theory of human cognition. REFERENCES Dijkstra, T. & van Heuven, W.J.B. (1998). The BIA model and bilingual word recognition. In J. Grainger & A.M. Jacobs (Eds.), Localist connectionist approaches to human cognition. Mahwah, NJ.: Erlbaum. Grainger, J. & Jacobs, A.M. (1998). On localist connectionism and psychological science. In J. Grainger & A.M. Jacobs (Eds.), Localist connectionist approaches to human cognition. Mahwah, NJ.: Erlbaum. Grainger, J. & Jacobs, A.M. (1998). Temporal integration of information in orthographic priming. Visual Cognition, in press. Green, CD. (1998) Are Connectionist Models Theories of Cognition? PSYCOLOQUY 9(4) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1998.volume.9/ psyc.98.9.04.connectionist-explanation.1.green Jacobs, A.M. & Grainger, J. (1992). Testing a semistochastic variant of the interactive activation model in different word recognition experiments. Journal of Experimental Psychology: Human Perception and Performance, 18, 1174-1188. Jacobs, A. M., & Grainger, J. (1994). Models of visual word recognition: Sampling the state of the art. Journal of Experimental Psychology: Human Perception and Performance, 20, 1311-1334. Jacobs, A.M., Rey, A., Ziegler, J.C, & Grainger, J. (1998). MROM-P: An interactive activation, multiple read-out model of orthographic and phonological processes in visual word recognition. In J. Grainger & A.M. Jacobs (Eds.), Localist connectionist approaches to human cognition. Mahwah, NJ.: Erlbaum. McClelland, J. L. & Rumelhart, D. E. (1981). An interactive activation model of context effects in letter perception: Part I. An account of basic findings. Psychological Review, 88, 375-407. Page, M. & Norris, D. (1998). Modeling immediate serial recall with a localist implementation of the primacy model. In J. Grainger & A.M. Jacobs (Eds.), Localist connectionist approaches to human cognition. Mahwah, NJ.: Erlbaum. Rumelhart, D.E., Hinton, G.E. & Williams, R.J. (1986). Learning internal represenatations by error propagation. In D.E. Rumelhart, J.L. McClelland, & the PDP research group, Parallel distributed processing: Explorations in the microstructure of cognition (Vol. 1). Cambridge, MA: Bradford Books. INSTRUCTIONS FOR PSYCOLOQUY 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 50,000. 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. Accepted PSYCOLOQUY target articles have been judged by 5-8 referees to be appropriate for Open Peer Commentary, the special service provided by PSYCOLOQUY to investigators in psychology, neuroscience, behavioral biology, cognitive sciences and philosophy who wish to solicit multiple responses from an international group of fellow specialists within and across these disciplines to a particularly significant and controversial piece of work. If you feel that you can contribute substantive criticism, interpretation, elaboration or pertinent complementary or supplementary material on a PSYCOLOQUY target article, you are invited to submit a formal electronic commentary. 1. Before preparing your commentary, please examine recent numbers of PSYCOLOQUY if not familiar with the journal. 2. Commentaries should preferably be up to ~200 lines (~1800 words) 3. Please provide a title for your commentary. 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 commentary should also have a brief (~100 word) abstract 4. All paragraphs should be numbered consecutively. Line length should not exceed 72 characters. The commentary 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, preparation and editing. Target article length should preferably be up to 1200 lines [c. 10,000 words]. All target articles, commentaries and responses must have (1) a short abstract (up to 200 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 URLs for retrieving full texts of target articles: http://cogsci.soton.ac.uk/psyc http://www.princeton.edu/~harnad/psyc.html gopher://gopher.princeton.edu:70/11/.libraries/.pujournals ftp://ftp.princeton.edu/pub/harnad/Psycoloquy ftp://cogsci.soton.ac.uk/pub/harnad/Psycoloquy news:sci.psychology.journals.psycoloquy Anonymous ftp archive is DIRECTORY pub/harnad/Psycoloquy HOST ftp.princeton.edu From aapo at myelin.hut.fi Tue Apr 28 07:57:08 1998 From: aapo at myelin.hut.fi (Aapo Hyvarinen) Date: Tue, 28 Apr 1998 14:57:08 +0300 Subject: ICA'99 call for papers Message-ID: <199804281157.OAA06512@myelin.hut.fi> -- We apologize if you receive multiple copies of this message. First Call for Papers: ------------- I C A ' 9 9 ------------- International Workshop on INDEPENDENT COMPONENT ANALYSIS and BLIND SIGNAL SEPARATION January 11-15, 1999 Aussois, France http://sig.enst.fr/~ica99 Submission deadline: July 15, 1998 ---------------------------------------------------------------------------- SCOPE ---------------------------------------------------------------------------- The workshop is devoted to recent advances in Independent Component Analysis and Blind Separation of Signals. It is intended to bring together researchers from the fields of artificial neural networks, signal processing, statistics, data analysis and all other domains connected to information processing. We are soliciting contributions covering all aspects of ICA and BSS: theory, methods, implementation and recent experimental results. The workshop will feature poster and oral presentations (no parallel sessions) and many opportunities for exchanges and informal discussions. ---------------------------------------------------------------------------- SPECIAL SESSIONS ---------------------------------------------------------------------------- Three special sessions will be organized on ICA applications: T. Sejnowski, Salk Institute, USA: Biomedical applications K. Torkkola, Motorola, Phoenix, USA: Speech and audio applications Y. Deville, UPS, Toulouse, France: General applications of ICA and BSS ---------------------------------------------------------------------------- VENUE ---------------------------------------------------------------------------- This one-week workshop will be held in Aussois, a small ski resort (alt. 1500m) in the magnificent mountains of La Vanoise in the heart of the French Alpes. The venue offers all the workshop facilities and we expect an enjoyable and productive `workshop atmosphere'. ---------------------------------------------------------------------------- SUBMISSION and PUBLICATION ---------------------------------------------------------------------------- Submission information will be available from our web site: http://sig.enst.fr/~ica99 Important dates: July 15, 1998 Submission of *full* paper Sep. 30, 1998 Notification of acceptance Jan. 11-15, 1999 Workshop All papers presented at the workshop will be collected in a volume of proceedings, which will be distributed to the participants on site. ---------------------------------------------------------------------------- SCIENTIFIC COMMITTEE ---------------------------------------------------------------------------- Shun-ichi Amari Brain Science Institute, RIKEN, Japan Tony Bell Salk Institute, USA Andrzej Cichocki Brain Science Institute, RIKEN, Japan Pierre Comon Eurecom, France Gustave Deco Siemens Research, Germany. Lieven De Lathauwer Katholieke Universiteit Leuven, Belgium Colin Fyfe University of Paisley, UK Simon Godsill University of Cambridge, UK Jean-Louis Lacoume Institut Nat. Polytechnique de Grenoble, France Ruey-Wen Liu University of Notre Dame, USA Odile Macchi CNRS/LSS, France Jean-Pierre Nadal Ecole Normale Superieure, Paris, France Erkki Oja Helsinki University of Technology, Finland Dinh-Tuan Pham CNRS/IMAG, France Jitendra Tugnait Auburn University, USA ---------------------------------------------------------------------------- ORGANIZATION ---------------------------------------------------------------------------- Organizers: Jean-Francois Cardoso CNRS and ENST, Paris, France Christian Jutten Institut Nat. Polytechnique de Grenoble, France Philippe Loubaton Universite de la Marne la Vallee, France Publicity: Aapo Hyvarinen Helsinki University of Technology, Finland Lieven De Lathauwer Katholieke Universiteit Leuven, Belgium ---------------------------------------------------------------------------- CONTACT INFORMATION ---------------------------------------------------------------------------- For more information contact: Jean-Francois Cardoso, ENST/SIG, 46 rue Barrault F-75634 Paris Cedex 13, France Internet: web site http://sig.enst.fr/~ica99 email ica99 at sig.enst.fr ---------------------------------------------------------------------------- From oby at cs.tu-berlin.de Wed Apr 29 06:37:47 1998 From: oby at cs.tu-berlin.de (Klaus Obermayer) Date: Wed, 29 Apr 1998 12:37:47 +0200 (MET DST) Subject: preprints available Message-ID: <199804291037.MAA16314@pollux.cs.tu-berlin.de> Dear Connectionists, I am happy to announce a series of papers on topographic clustering, self-organizing maps, dissimilarity data, and kernels. Cheers Klaus ------------------------------------------------------------------------ Prof. Klaus Obermayer phone: 49-30-314-73442 FR2-1, NI, Informatik 49-30-314-73120 Technische Universitaet Berlin fax: 49-30-314-73121 Franklinstrasse 28/29 e-mail: oby at cs.tu-berlin.de 10587 Berlin, Germany http://ni.cs.tu-berlin.de/ ========================================================================= A Stochastic Self-organizing Map for Proximity Data T. Graepel and K. Obermayer We derive an efficient algorithm for topographic mapping of proximity data (TMP), which can be seen as an extension of Kohonen's Self- Organizing Map to arbitrary distance measures. The TMP cost function is derived in a Baysian framework of Folded Markov Chains for the description of autoencoders. It incorporates the data via a dissimilarity matrix ${\mathcal D}$ and the topographic neighborhood via a matrix ${\mathcal H}$ of transition probabilities. From the principle of Maximum Entropy a non-factorizing Gibbs-distribution is obtained, which is approximated in a mean-field fashion. This allows for Maximum Likelihood estimation using an EM-algorithm. In analogy to the transition from Topographic Vector Quantization (TVQ) to the Self-organizing Map (SOM) we suggest an approximation to TMP which is computationally more efficient. In order to prevent convergence to local minima, an annealing scheme in the temperature parameter is introduced, for which the critical temperature of the first phase-transition is calculated in terms of ${\mathcal D}$ and ${\mathcal H}$. Numerical results demonstrate the working of the algorithm and confirm the analytical results. Finally, the algorithm is used to generate a connection map of areas of the cat's cerebral cortex. to appear in: Neural Computation preprint: http://ni.cs.tu-berlin.de/publications/#journals ------------------------------------------------------------------------- Fuzzy Topographic Kernel Clustering} T. Graepel and K. Obermayer A new topographic clustering algorithm is proposed, which - by the use of integral operator kernel functions - efficiently estimates the centers of clusters in high-dimensional feature spaces, which is related to data space by some nonlinear map. Like in the Self-Organizing Map topography is imposed by assuming finite transition probabilities between cluster indices. The optimization of the associated cost function is achieved by estimating the parameters via an EM-scheme and deterministic annealing. The effect of different radial basis function kernels on topographic maps of handwritten digit data is examined in computer simulations. In: W. Brauer, editor, Proceedings of the 5th GI Workshop Fuzzy Neuro Systems '98, pages 90-97, 1998. preprint: http://ni.cs.tu-berlin.de/publications/#conference ------------------------------------------------------------------------- An Annealed Self-Organizing Map for Source-Channel Coding M. Burger, T. Graepel, and K. Obermayer We derive and analyse robust optimization schemes for noisy vector quantization on the basis of deterministic annealing. Starting from a cost function for central clustering that incorporates distortions from channel noise we develop a soft topographic vector quantization algorithm (STVQ) which is based on the maximum entropy principle and which performs a maximum-likelihood estimate in an expectation-maximization (EM) fashion. Annealing in the temperature paramete $\beta$ leads to phase transitions in the existing code vector representation during the cooling process for which we calculate critical temperatures and modes as a function of eigenvectors and eigenvalues of the covariance matrix of the data and the transition matrix of the channel noise. A whole family of vector quantization algorithms is derived from STVQ, among them a deterministic annealing scheme for Kohonen's self-organizing map (SOM). This algorithm, which we call SSOM, is then applied to vector quantization of image data to be sent via a noisy binary symmetric channel. The algorithm's performance is compared to those of LBG and STVQ. While it is naturally superior to LBG, which does not take into account channel noise, its results compare very well to those of STVQ, which is computationally much more demanding. to appear in: NIPS 10 proceedings preprint: http://ni.cs.tu-berlin.de/publications/#conference The theory is quite well described in: T. Graepel, M. Burger, and K. Obermayer. Phase transitions in Stochastic Self-Organizing Maps. Phys. Rev. E, 56(4):3876-3890, 1997. preprint: http://ni.cs.tu-berlin.de/publications/#journals ----------------------------------------------------------------------- Review-style paper for the practioneers: Self-Organizing Maps: Generalization and New Optimization Techniques T. Graepel, M. Burger, and K. Obermayer We offer three algorithms for the generation of topographic mappings to the practitioner of unsupervised data analysis. The algorithms are each based on the minimization of a cost function which is performed using an EM algorithm and deterministic annealing. The soft topographic vector quantization algorithm (STVQ) -- like the original Self-Organizing Map (SOM) -- provides a tool for the creation of self-organizing maps of Euclidean data. Its optimization scheme, however, offers an alternative to the heuristic stepwise shrinking of the neighborhood width in the SOM and makes it possible to use a fixed neighborhood function solely to encode desired neighborhood relations between nodes. The kernel-based soft topographic mapping (STMK) is a generalization of STVQ and introduces new distance measures in data space based on kernel functions. Using the new distance measures corresponds to performing the STVQ in a high-dimensional feature space, which is related to data space by a nonlinear mapping. This preprocessing can reveal structure of the data which may go unnoticed if the STVQ is performed in the standard Euclidean space. The soft topographic mapping for proximity data (STMP) is another generalization of STVQ that enables the user to generate topographic maps for data which are given in terms of pairwise proximities. It thus offers a flexible alternative to multidimensional scaling methods and opens up a new range of applications for Self-Organizing Maps. Both STMK and STMP share the robust optimization properties of STVQ due to the application of deterministic annealing. In our contribution we discuss the algorithms together with their implementation and provide detailed pseudo-code and explanations. to appear in: Neurocomputing preprint: http://ni.cs.tu-berlin.de/publications/#journals From annesp at vaxsa.csied.unisa.it Wed Apr 29 07:05:10 1998 From: annesp at vaxsa.csied.unisa.it (annesp@vaxsa.csied.unisa.it) Date: Wed, 29 Apr 1998 12:05:10 +0100 Subject: E.R.CAIANIELLO SUMMER SCHOOL DEADLINE Message-ID: <98042912051019@vaxsa.csied.unisa.it> ***************************************************************** Please post **************************************************************** International Summer School ``Neural Nets E. R. Caianiello" 3rd Course "A Course on Speech Processing, Recognition, and Artificial Neural Networks" web page: http://wsfalco.ing.uniroma1.it/Speeschool.html The school is jointly organized by: INTERNATIONAL INSTITUTE FOR ADVANCED SCIENTIFIC STUDIES (IIASS) Vietri sul Mare (SA) Italy, ETTORE MAJORANA FOUNDATION AND CENTER FOR SCIENTIFIC CULTURE (EMFCSC) Erice (TR), Italy Supported by: EUROPEAN SPEECH COMMUNICATION ASSOCIATION (ESCA) Sponsored by: SALERNO UNIVERSITY, Dipartimento di Scienze Fisiche E.R. Caianiello (Italy) DIRECTORS OF THE COURSE DIRECTORS OF THE SCHOOL AND ORGANIZING COMMITTEE: Gerard Chollet (France). Maria Marinaro (Italy) M. Gabriella Di Benedetto (Italy) Michael Jordan (USA) Anna Esposito (Italy) Maria Marinaro (Italy) PLACE: International Institute for Advanced Scientific Studies (IIASS) Via Pellegrino 19, 84019 Vietri sul Mare, Salerno (Italy) DATES: 5th-14th October 1998 POETIC TOUCH Vietri (from "Veteri", its ancient Roman name) sul Mare ("on sea") is located within walking distance from Salerno and marks the beginning of the Amalfi coast. Short rides take to Positano, Sorrento, Pompei, Herculaneum, Paestum, Vesuvius, or by boat, the islands of Capri, Ischia, and Procida. Velia (the ancient "Elea" of Zeno and Parmenide) is a hundred kilometers farther down along the coast. Student Fee: 1500 dollars Student fee include accommodations (arranged by the school), meals, one day of excursion, and a copy of the proceedings of the school. Transportation is not included. A few scholarships are available for students who are otherwise unable to participate at the school, and who cannot apply for the grants offered by ESCA. The scholarship will partially cover lodging and living expenses. Day time: 3 hour in the morning, three hour in the afternoon. Day free: One day with an excursion of the places around. AIMS: The aim of this school is to present the experiments, the theories and the perspectives of acoustic phonetics, as well as to discuss recent results in the speech literature. The school aims to provide a background for further study in many of the fields related to speech science and linguistics, including automatic speech recognition. The school will bring together leading researchers and selected students in the field of speech science and technology to discuss and disseminate the latest techniques. The school is devoted to an international audience and in particular to all students and scientists who are working on some aspects of speech and want to learn other aspects of this discipline. MAJOR TOPICS The school will cover a number of broad themes relevant to speech, among them: 1) Speech production and acoustic phonetics 2) Articulatory, acoustic, and prosodic features 3) Acoustic cues in speech perception 4) Models of speech perception 5) Speech processing (Preprocessing algorithms for Speech) 6) Neural Networks for automatic speech recognition 7) Multi-modal speech recognition and recognition in adverse environments. 8) Speech to speech translation (Vermobil and CSTAR projects) 9) Applications (Foreign Language training aids, aids for handicapped, ....). 10) Stochastic Models and Dialogue systems FORMAT The meeting will follow the usual format of tutorials and panel discussions together with poster sessions for contributed papers. The following tutorials are planned: ABEER ALWAN UCLA University (CA) USA "Models of Speech Production and Their Application in Coding and Recognition" ANDREA CALABRESE University of Connecticut (USA) "Prosodic and Phonological Aspects of Language" GERARD CHOLLET CNRS - ENST France "ALISP, Speaker Verification, Interactive Voice Servers" PIERO COSI CNR-Padova Italy "Auditory Modeling and Neural Networks" RENATO DE MORI Universite d' Avignon, France "Statistical Methods for Automatic Speech Recognition" M. GABRIELLA DI BENEDETTO Universita' degli Studi di Roma "La Sapienza", Rome, Italy ``Acoustic Analysis and Perception of Classes of Sounds (vowels and consonants)" BJORN GRANSTROM Royal Institute of Technology (KTH) Sweden "Multi-modal Speech Synthesis with Application" JEAN P. HATON Universite Henri-Poincare, CRIN-INRIA, France "Neural Networks for Automatic Speech Recognition" HYNEK HERMANSKY Oregon Graduate Institute, USA "Goals and Techniques of Speech Analysis" HERMANN NEY Computer Science Department, Aachen Germany "Algorithms for Large Vocabulary Speech Recognition" "Text to Speech Translation using Statistical Methods" JOHN OHALA University of California at Berkeley (CA) USA "Articulatory Constraints on Distinctive Features" JEAN SYLVAIN LIENARD LIMSI-CNRS, France "Speech Perception, Voice Perception" "Beyond Pattern Recognition" PROCEEDINGS The proceedings will be published in the form of a book containing tutorial chapters written by the lecturers and possibly shorter papers from other participants. One free copy of the book will be distributed to each participant. LANGUAGE The official language of the school will be English. POSTER SUBMISSION There will be a poster session for contributed presentations from participants. Proposals consisting of a one page abstract for review by the organizers should be submitted with applications. DURATION Participants are expected to arrive in time for the evening meal on Sunday 4th October and depart on Tuesday 15th October. Sessions will take place from Monday 5th-Wednesday 14th. COST The cost per participant of 1.500 $ dollars covers accommodation (in twin rooms), meals for the duration of the course, and one day of excursion. -- A supplement of 40 dollars per night should be paid for single room. Payment details will be notified with acceptance of applications. GRANTS -- A few ESCA grants are available for participants (which cover tuition and, maybe, part of the lodging). See http://ophale.icp.inpg.fr/esca/grants.html for further information. Individual applications for grants should be sent to Wolfgang Hess by e-mail: wgh at sunwgh.ikp.uni-bonn.de ELIGIBILITY The school is open to all suitably qualified scientists from around the world. APPLICATION PROCEDURE: Important Date: Application deadline: May 15 1998 Notification of acceptance: May 30 1998 Registration fee payment deadline: July 10 1998 People with few years of experience in the field should include a recommendation letter of their supervisor or group leader Places are limited to a maximum of 60 participants in addition to the lecturers. These will be allocated on a first come, first served basis. ************************************************************************** APPLICATION FORM Title:^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Family Name:^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Other Names:^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Name to appear on badge:^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Mailing Address (include institution or company name if appropriate): ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Phone:^^^^^^^^^^^^^^^^^^^^^^Fax:^^^^^^^^^^^^^^^^^^^ E-mail:^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Date of Arrival : Date of Departure: Will you be applying for a ESCA grant ? yes/no* *(please delete the alternatives which do not apply) Will you be applying for a scholarship ? yes/no* *(please delete the alternatives which do not apply) *(please include in your application a justification for scholarship request) ***************************************************************** Please send the application form together the recommendation letter by electronic mail to: iiass at tin.it, subject: summer school; or by fax: +39 89 761 189 (att.ne Prof. M. Marinaro) or by ordinary mail to the address below: IIASS Via Pellegrino 19, I84019 Vietri sul Mare (Sa) Italy For further information please contact: Anna Esposito International Institute for advanced Scientific Studies (IIASS) Via Pellegrino, 19, 84019 Vietri sul Mare (SA) Italy Fax: + 39 89 761189 e-mail: annesp at vaxsa.csied.unisa.it ================== RFC 822 Headers ==================