Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sdd.hp.com!spool.mu.edu!news.cs.indiana.edu!att!emory!ogicse!moe!maxwebb From: maxwebb@moe.cse.ogi.edu (Max G. Webb) Newsgroups: comp.ai.philosophy Subject: Re: Continuous vs discrete Message-ID: <19628@ogicse.ogi.edu> Date: 5 Apr 91 01:12:16 GMT Article-I.D.: ogicse.19628 References: <91082.223501DOCTORJ@SLACVM.SLAC.STANFORD.EDU> <1991Mar25.141743.21124@news.larc.nasa.gov> <19175@ogicse.ogi.edu> <19392@ogicse.ogi.edu> Sender: news@ogicse.ogi.edu Organization: Oregon Graduate Institute - Computer Science & Engineering Lines: 76 Here are some references: Koch and Segev editors, "Methods in Neuronal Modeling", copyright 1989, Massachusetts Institute of Technology This has a lot of work you would find interesting, the earlier articles address very detailed models of a _few_ neurons (as in tritonias central pattern generator for swimming) and less detailed models of more complex systems of human hearing. Also some good summaries of numerical methods. Chapter 7 Associative Network Models for Central Pattern Generators I originally saw the summary of the work in simulating the swimming behavior of the lamprey in the Inetmail.connectionists group, but didn't save it. Why don't you post a request there? If not, I'll go and do a keyword library search (but it costs me money, and you want the ref) They used a cray, simulated thousands of neurons and obtained a image of the swimming lamprey running at 1/10th normal speed! That is enough to keep you busy til I find the other one (central pattern generator for stomach ganglion of the lobster. (I do have a life apart from usenet - i am doing research on the olfactory cortex) In article magi@utu.fi (Marko Gronroos) writes: >> that these networks are simulable? > >Hmm. The differences may not cause too big errors in simple >simulations such as these, but may cause big errors in networks of >larger size and logical complexity. Possibly, but if these systems are so computationally nonrobust that rounding and discretization errors in a computer simulation obliterate their value, then how could they have evolved? Keep in mind that the neurons and the architecture of the nets have been changing and evolving at the same time? >Sorry again. Thought that you were talking about int's... But I'd say >that my example would have been good if you had been talking about int's. Actually, no. The tremendous range of light values detectable (1 photon up to a sunbeam) is compressed to a narrower range before the nervous system ever sees it, by the photoreceptor (first of all). Secondly, there is plenty of evidence that it is _edge_ information that is passed back, possibly other compressions of the data. It is NOT levels of illumination, as can be illustrated by numerous illusions. While i also like floating point, you would have a very hard time convincing a biologist that a change of 2^-24 in the operating levels of a neuron would destabilize the system! > > The question number 1 is that at what point of discretity would the >difference between the simulation and the nature become significant? >(With ANN's..) Well, the numerical techniques we are using were not invented for this problem; error analysis has been around a long time. Why don't you just get a book on it out of the library? (the book i mentioned has a little on this) >In those systems that you mentioned simulations try to simulate real >physical behaviour. Most current artificial neural networks simulate >the presumed logical behaviour of some objects that we really >don't know much about. > It's not only important that one studies, it's also important _what_ >one studies. Can you give me some hints? :-) Well, heres a couple, Koch and Segev, and any good Numerical Analysis book. >Marko Gronroos ! Tel. +358-21-445613 ! Max