Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!usc!apple!agate!eos!ptolemy!laird
From: laird@ptolemy.arc.nasa.gov (Philip Laird)
Newsgroups: comp.ai.neural-nets
Subject: Re: When does a hopfield net converge ?
Message-ID: <1322@rayleigh.ptolemy.arc.nasa.gov>
Date: 11 Jul 89 20:37:58 GMT
References: <6009@uklirb.UUCP> <3738@cps3xx.UUCP> <3376@ncsuvx.ncsu.edu>
Reply-To: laird@pluto.arc.nasa.gov
Organization: NASA Ames Research Center, Moffett Field, CA
Lines: 17

In article <3376@ncsuvx.ncsu.edu> thomae@cslvs5.UUCP (Doug Thomae) writes:
>... I have heard mention of more general theorem that show that a Hopfield
>network will also converge under some conditions when the connections
>are not symmetric, but I don't know the reference for it.

Convergence (or lack thereof) is rather well understood theoretically.
Two recent references are:

Bruck and Sanz, Int. Journal of Intelligent systems, 3, p. 59-75, 1988.

Bruck and Goodman, IEEE Trans. on Information Theory, IT-34 (Sept. 88).



-- 
Phil Laird, NASA Ames Research Center, Moffett Field, CA 94035
(415)-694-3362  LAIRD@PLUTO.ARC.NASA.GOV