ronl@tessi.UUCP (Ron Lunde) (05/09/91)
Hi! I'm trying to write a fairly general neural network simulator using C++, so I can experiment with different types of network models. I would like to write it in such a way that I can output an ASCII description of the network, which can be easily ported to special-purpose NN hardware or to other simulators. In connection with this, I have a few questions for you, the neural network world: 1) Is there a standard interchange format for neural networks? If so, or almost, could you send me references? If not, do you think an interchange format would be useful, and would you like to help develop one? For example, the nodes and connection weights might be specified in a simple table: node node weight ------------------------------- 1 2 .87 1 3 -0.4 ... ... ... If the answer to this is "no standard exists" and "one would be useful", I will try to help find someone to coordinate it unless someone else already is working on it. 2) Has anyone done a class analysis of neural network varieties? For instance, given a list of neural network models and algorithms: Perceptron Back Propagation Feed Forward Hopfield Bayesian Adaptive Resonance Bidirectional Associative Memory Kohonen Competitive Learning ... Has anyone organized this into a class hierarchy (with multiple inheritance, if necessary)? If so, would someone please send me some references? 3) Would anyone be willing to look over my class definitions and methods (C++) and give me some feedback on how I've broken it down so far? Thanks! ----------------------------------------------------------------------------- Ron E. Lunde ronl@tessi sun.com!nosun!tessi!ronl Test Systems Strategies, Inc. 8205 S.W. Creekside Place Beaverton, OR 97005 USA Phone: (503) 643-9281 FAX: (503) 646-4954 -----------------------------------------------------------------------------