Neural networks are a buzz word. Why ? They are a very powerful tool in non-linear statistical analysis. As such they have found their way into many fields - control theory, natural language processing, image processing, process modeling - and are strongly supported by industry. There is a lot of up-to-date information available on the WWW and we have listed important web sites which will open the doors for you to a very exciting field of research.
We aim to give computational support for the general development of neural network algorithms and applications. The following lists packages which have been installed (or can be installed upon request) at the Laboratoire National Saturne, France.
ART = Adaptive Resonance Theory dLVQ = Dynamic Learning Vector Quantizer BP = Backpropagation CC = Cascade Correlation LVQ = Learning Vector Quantizer SOM = Self-Organizing Map HAM = Hamming Net HOP = Hopfield Network M3 = Madaline III MFT = Mean Field Theory OLAM= Optimal Linear Assoc. Memory QP = QuickPropagation RBP = Recurrent Backprop RCE = Restricted Coulomb Energy
NeuDL is a tool with an interpreted programming language interface to build, train, test, and run neural network designs. The functionality is limited to backpropagation trained neural networks. The tool is written in C++ and includes a NeuDL-to-C++ translater.
VieNet2 is a toolkit for implementing simulations of neural networks without need to take care of memory allocation, pointer-handling and graphical representation. It comprises a set of functions and datatypes, written in ANSI-C, which will be linked with your own main-program. VieNet2 is designed to support your development of neural networks as much as possible while avoiding restrictions other simulation tools will have. With the availability of the full sourcecode you are enabled to combine nearly every C-program you want with a neural network. You may start with the graphical interface, which is generated automatically by VieNet2, for testing, training and development of your network. Later on you may enhance the network with your own learning functions or combine it with your hardware drivers for data requisition or control.
The software consists of a code generator that builds neural network simulations by reading a network description (written in a language called "Aspirin") and generates an ANSI C simulation. An interface (called "MIGRAINES") is provided to export data from the neural network to visualization tools. (see also FAQ)
A luxurious simulator for many types of nets with X11 interface. University of Stuttgart, Germany (see also FAQ)
XERION is a neural network simulator from Drew van Camp at the University of Toronto. It provides a library of routines for building networks and graphically displaying them. (see also FAQ)
This is a prototype that stems from an ESPRIT project.
ICANN95, AIHENP95-1 : Application of Neural Network Algorithms for Data Analysis in a High Energy Physics Experiment (in PostScript : ICANN, AIHENP). The simulation and reconstruction of calorimeter data for an experiment at the Laboratoire National Saturne gives an example of how and where these algorithms are applied.
Please send e-mail to fuchs@republique.saclay.cea.fr or FAX your request to +33 1 69.08.90.11 for information on the ETA project. If you have problems or comments concerning our WWW service, please send e-mail to the same address.
This page, and all contents, are Copyright (C) 1995 by Paul Fuchs, Laboratoire National Saturne, 91191 Gif-sur-Yvette Cedex, France under the GNU public software license agreement version 2 or later.
Back to the ETA Home Page