Neural Network Simulation Classes for the NeXT Computer By: Ralph Zazula University of Arizona zazula@bonehead.tucson.az.us (NeXT Mail) Jan 10, 1992 This package contains Classes that can be used to create software simulations of neural network architectures. These were written by me for use in my research with the goal of rapid creation of code to explore various network topologies and learning algorithms. I would like to hear input from people using these ± especially if you have found errors or have improvements ± or just to hear what NeXTy things you are doing in your neural-network research. You cannot use these for commercial purposes. The classes contained are: Neuron Implements a number of different types of neurons BackPropEngine Implements a back-propagation network The file bptest.m shows an example of using the BackPropEngine class to learn the identity function. The file boltz.m uses a fully-connected boltzman machine to learn the XOR problem. Take note of the SYMMETRIC definition in Neuron.h. I will be making a cleaner interface to the weights in the near future but for now you must change the define accordingly. Most everything else is documented through the Class descriptions, code or example programs. If you have any questions/comments, feel free to e-mail me. Ralph Zazula
These are the contents of the former NiCE NeXT User Group NeXTSTEP/OpenStep software archive, currently hosted by Netfuture.ch.
