Technical Report Series in Neural and Computational Learning

[John Shawe-Taylor]

The European Community ESPRIT Working Group in Neural and Computational 
Learning Theory (NeuroCOLT) has produced a set of new Technical Reports
available from the remote ftp site described below. They cover topics in
real valued complexity theory, computational learning theory, and analysis
of the computational power of continuous neural networks.  Abstracts are
included for the titles.

*** Please note that the location of the files was changed at the beginning of
** the year, so that any copies you have of the previous instructions should be 
* discarded. The new location and instructions are given at the end of the list.


----------------------------------------
NeuroCOLT Technical Report NC-TR-96-043:
----------------------------------------
Elimination of Constants from Machines over Algebraically Closed Fields
by  Pascal Koiran, Ecole Normale Sup\'erieure de Lyon, France

Abstract:
Let $\k$ be an algebraically closed field of characteristic 0.  We show
that constants can be removed efficiently from any machine over $\k$
solving a problem which is definable without constants.  This gives a
new proof of the transfer theorem of Blum, Cucker, Shub \& Smale for
the problem $\p \stackrel{?}{=}\np$.  We have similar results in
positive characteristic for non-uniform complexity classes.  We also
construct explicit and correct test sequences (in the sense of Heintz
and Schnorr) for the class of polynomials which are easy to compute.


----------------------------------------
NeuroCOLT Technical Report NC-TR-96-044:
----------------------------------------
Hilbert's Nullstellensatz is in the Polynomial Hierarchy
by  Pascal Koiran, Ecole Normale Sup\'erieure de Lyon, France

Abstract:
We show that if the Generalized Riemann Hypothesis is true, the problem
of deciding whether a system of polynomial equations in several complex
variables has a solution is in the second level of the polynomial
hierarchy. The best previous bound was PSPACE. The possibility that
this problem might be NP-complete is also discussed (it is well-known
to be NP-hard).


----------------------------------------
NeuroCOLT Technical Report NC-TR-96-045:
----------------------------------------
Networks of Spiking Neurons: The Third Generation of Neural Network Models
by  Wolfgang Maass, Technische Universitaet Graz, Austria

Abstract:
The computational power of formal models for networks of spiking
neurons is compared with that of other neural network models based on
McCulloch Pitts neurons (i.e. threshold gates) respectively sigmoidal
gates. In particular it is shown that networks of spiking neurons are
computationally more powerful than these other neural network models. A
concrete biologically relevant function is exhibited which can be
computed by a single spiking neuron (for biologically reasonable values
of its parameters), but which requires hundreds of hidden units on a
sigmoidal neural net.
This article does not assume prior knowledge about spiking neurons, and
it  contains an extensive list of references to the currently available
literature on computations in networks of spiking neurons and relevant
results from neurobiology.


----------------------------------------
NeuroCOLT Technical Report NC-TR-96-046:
----------------------------------------
Use Of Neural Network Ensembles for Portfolio Selection and Risk Management
by  D.L.Toulson, Intelligent Financial Systems Ltd., UK
    S.P.Toulson, London School Of Economics, UK

Abstract:
A well known method of managing the risk whilst maximising the return
of a portfolio is through Markowitz Analysis of the efficient set. A
key pre-requisite for this technique is the accurate estimation of the
future expected returns and risks (variance of re turns) of the
securities contained in the portfolio along with their expected
correlations. The estimates for future returns are typically obtained
using weighted averages of historical returns of the securities
involved or other (linear) techniques. Estimates for the volatilities
of the securities may be made in the same way or through the use of
(G)ARCH or stochastic volatility (SV) techniques.
In this paper we propose the use of neural networks to estimate future
returns and risks of securities. The networks are arranged into {\em
committees}. Each committee contains a number of independ ently trained
neural networks. The task of each committee is to estimate either the
future return or risk of a particular security.  The inputs to the
networks of the committee make use of a novel discriminant analysis
technique we have called {\em Fuzzy Discriminants Analysis}.
The estimates of future returns and risks provided by the committees
are then used to manage a portfolio of 40 UK equities over a five year
period (1989-1994). The management of the portfolio is constrained such
that at any time it should have the same risk characteristic as the
FTSE-100 index. Within this constraint, the portfolio is chosen to
provide the maximum possible return. We show that the managed portfolio
significantly outper forms the FTSE-100 index in terms of both overall
return and volatility.


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

***************** ACCESS INSTRUCTIONS ******************

The Report NC-TR-96-001 can be accessed and printed as follows 

% ftp ftp.dcs.rhbnc.ac.uk  (134.219.96.1)
Name: anonymous
password: your full email address
ftp> cd pub/neurocolt/tech_reports
ftp> binary
ftp> get nc-tr-96-001.ps.Z
ftp> bye
% zcat nc-tr-96-001.ps.Z | lpr -l

Similarly for the other technical reports.

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

In some cases there are two files available, for example,
nc-tr-96-002-title.ps.Z
nc-tr-96-002-body.ps.Z
The first contains the title page while the second contains the body 
of the report. The single command,
ftp> mget nc-tr-96-002*
will prompt you for the files you require.

A full list of the currently available Technical Reports in the 
Series is held in a file `abstracts' in the same directory.

The files may also be accessed via WWW starting from the NeuroCOLT 
homepage (note that this is undergoing some corrections and may be 
temporarily inaccessible):

http://www.dcs.rhbnc.ac.uk/neural/neurocolt.html


Best wishes
John Shawe-Taylor