Path: utzoo!attcan!uunet!wuarchive!mit-eddie!uw-beaver!ubc-cs!fornax!mcguire
From: mcguire@fornax.UUCP (Michael McGuire)
Newsgroups: comp.ai.neural-nets
Subject: Multi-layer Perceptron Network Design
Keywords: MLP neural network
Message-ID: <2051@fornax.UUCP>
Date: 31 Jan 91 04:52:38 GMT
Distribution: na
Organization: School of Computing Science, SFU, Burnaby, B.C. Canada
Lines: 7

Does anyone know of any books or papers that describe in detail how
one would go abount designing a multi-layer perceptron neural network
given a certain problem. I am interested in the selection of layers,
number of hidden units, learning rate etc. Most of the literature seems to
use a trial-and-error method or experience. Given a problem, and a set of
patterns to analyze, is there a good way to determine whether a certain
network could learning the mapping. Much appreciated.