Newsgroups: comp.ai.neural-nets,comp.ai.alife,comp.theory.cell-automata,comp.theory.dynamic-sys,comp.theory.self-org-sys
Path: cantaloupe.srv.cs.cmu.edu!rochester!cornellcs!newsstand.cit.cornell.edu!portc01.blue.aol.com!news-peer.gsl.net!news.gsl.net!news.sgi.com!www.nntp.primenet.com!nntp.primenet.com!newsfeed1.aimnet.com!news.netserv.com!pagesat.net!NEWS!not-for-mail
From: jemyers@oregoncoast.com (John Myers)
Subject: Pattern recognition using a random Boolean network
Message-ID: <a77cc$f261.284@NEWS>
Date: Mon, 07 Oct 1996 23:38:01 GMT
X-Newsreader: WinVN 0.99.7
MIME-Version: 1.0
Content-Type: Text/Plain; charset=US-ASCII
Lines: 23
Xref: glinda.oz.cs.cmu.edu comp.ai.neural-nets:33888 comp.ai.alife:6570 comp.theory.cell-automata:4572 comp.theory.dynamic-sys:2377 comp.theory.self-org-sys:1191

Recently I've been experimenting with using a random Boolean network --  RBN -- 
(also known as disordered cellular automata or Kauffman net) to solve very 
simple pattern recognition problems, such as sorting fruit on a conveyor belt 
based on various sensor inputs. I got started on this after reading a paper by 
Andy Wuensche at SFI in which he proposes RBNs as a model of memory and 
learning in the brain.  The parameters of the network ("wiring" and transition 
rules) determine the network dynamics (deterministic), thereby organizing all 
possible "states" of the network into basins of attraction, which can be 
interpreted as memory categories (so can tree structures within the basins) .  

In my application, briefly, I map a "training set" of inputs onto binary 
strings,select input/output pairs to sort the input data into desired 
categories, then determine the corresponding network parameters (least- 
connected) using an algorithm I developed (an "inverse" algorithm which has 
other uses also). This approach may offer promise, but it is too early to say. 
Initial results are at least interesting.
 
I would like to know if anyone out there is using RBNs, or has seen them used 
to solve pattern recognition problems, or has a great interest in the subject. 
I'm looking for references and/or dialogue on the subject.

Thanks.


