Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ukma!rutgers!njin!hubey From: hubey@pilot.njin.net (Hubey) Newsgroups: comp.ai.neural-nets Subject: derivatives in neural nets Keywords: derivatives, nets, neural Message-ID: Date: 30 Apr 89 03:42:40 GMT Organization: NJ InterCampus Network, New Brunswick, N.J. Lines: 23 I am not familiar with work done with neural networks other than simple knowledge that the "neurons" are connected to other neurons and that the rules for a neuron firing is that if the "weighted average" of the inputs to a neuron exceeds a thresh hold value it fires. My question is; Have there been any models where derivatives are employed. I remember reading years ago (from experiments on frog brains) that certain neurons only fire when particular patterns visual patterns occur. Still other neurons seem to fire periodically once "set off"-- not just once. It seems to me that this kind of phenomena are asking for slightly more than the "cross-the-threshold & fire once" type of models. In any case, I don't know where to start reading. If anyone knows of different types of neural net models, I'd be grateful for the info. Many thanks ... mark -- H.M. Hubey VOICE: 201-893-5269 hubey@OSultrix.montclair.edu hubey@pilot.njin.net hubey@apollo.montclair.edu