Neural Network Simulation Programs

There are now literally hundreds of neural network simulation programs available commercially, and as shareware and freeware. See, for example, listings of NNW simulators at Neuronet, DSI, and PNL shareware and commercial ware. Most offer back-propagation but many have a wide selection of other learning algorithms. Below is a short list of some simulators that might be of interest to HEP researchers. See also the SLAC - High Energy Physics Software.

AND HNet
BrainMaker: Description: IBM PC program with Back-propagation. Also, emulates the ETANN chip and a version comes with the iNNTS ETANN development system . They also have their own PC accelerator card. In addition, BrainMaker now runs on the Adaptive Solutions CNAPS PC cards.
DynaMind: Description: IBM PC program with several algorithms including Back-propagation and Madeline III. Also, emulates the ETANN chip and iDynamind version comes with the iNNTS ETANN development system .
JETNET 3.0: Description: A Fortran neural network program (latest release Dec. 1993). Specifically aimed at HEP community. Will run on virtually any machine with a F77 compiler.
NEURAL 2.00: Description: A program for neural net and statistical pattern recognition by R. Odorico, Univ. Bologna. Offers Fortran 77 coded program with BP, LVQ2, and Fisher Discriminates algorithms and emulation of Intel ETANN and Siemens MA-16 chips. See the paper, a program file, and a test-run.
Logical Designs THINKS: Description: "A Windows neural network program with a number of learning algorthms including: Back Propagation, Quick Propagation, Jacob's Enhanced Back Propagation, Recurrent Back Propagation, Kohonen Winner Take All, Kohonen Learning Vector Quantization, Simulated Annealing, and Cascade Correlation Learning. (More algorithms will come with the THINKS Professional version due out in early 1995.) All network calculations are in 32-bit double precision. Comes with easy to use GUI, allowing numerous network and parameter variations. A "test while training" option helps to prevent overtraining by showing network error on the test set during the training. Other features include preprocessing functions, various transfer function options, five error function options, a Run-Time library, a Windows DLL, etc."
Mathematica and Neural Networks: Description: Mathematica doesn't have an explicit package on it but there is a book on implementing neural networks with Mathematica. See James A. Freeman: Simulating Neural Networks with Mathematica , Addison-Wesley, 1994.
MathWorks MATLAB Neural Network Toolbox: Description: The Matlab mathematical analysis package now has a neural network toolbox available. Version 2.0 features several learning algorithms and architectures including: backpropagation, Levenberg-Marquardt fast backprop, radial basis functions, adaptive linear networks, Elman recurrent neetworks, LVQ, etc. Tutorials and examples of each type of network are given. The extensive graphing capabilities of Matlab can be incorporated into the network analyses. There are no limits on network size or connectivity. Other matlab toolboxes such as the signal and image processing toolboxes can be combined with the neural net toolbox to create complex multi-modular simulations.
NeuralWorks Professional II/Plus: Description: "A comprehensive neural network training, refining, and deployment development tool. Features include Logicon Projection Network, Modular Neural Network, Fuzzy Art Map, Radial Basis Functions, Quick-Prop, General Regression Neural Network, and update versions of Cascade Correlation and Probablistic Neural Network. The software includes a Train-Test option that optimizes neural network performance.:", from AI Experts, 1993 Neural Resource Guide.
NeuroSolutions for Windows: Description: "NeuroSolutions is a graphical neural network simulation tool. It is the only commercial Windows package to support trajectory learning with backpropagation through time. Because of its object-oriented design, NeuroSolutions provides the flexibility needed to construct a wide range of learning paradigms and network topologies. Its GUI and extensive probing ability streamline the experimentation process by providing real-time analysis of the network during learning. Construct any neural network belonging to the additive model, including locally and globally recurrent systems. Use a variety of unsupervised learning procedures, such as Hebbian, Sanger's, Oja's, Competitive and Kohonen. Implement RBF, PCA, counterpropagation and other hybrid network topologies by seamlessly integrating supervised and unsupervised learning."
PDP++ Software: Description: PDP++ software is a new neural-network simulation system written in C++. It represents the next generation of the PDP software released with the McClelland and Rumelhart "Explorations in Parallel Distributed Processing Handbook", MIT Press, 1987. It is easy enough for novice users, but very powerful and flexible for research use.

NNW/HEP Home Page

Authors: Clark S. Lindsey , Bruce Denby , & Thomas Lindblad

Curator: Clark S. Lindsey (lindsey@particle.kth.se)
Latest revision: 31 Dec 98