home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.theory.cell-automata
- Path: sparky!uunet!paladin.american.edu!gatech!enterpoop.mit.edu!bloom-beacon!INTERNET!dont-send-mail-to-path-lines
- From: rucker@sjsumcs.SJSU.EDU (Rudy Rucker)
- Subject: evolving cas
- Message-ID: <199301012145.AA20860@sjsumcs.SJSU.EDU>
- Sender: daemon@athena.mit.edu (Mr Background)
- Organization: The Internet
- Distribution: inet
- Date: Fri, 1 Jan 1993 21:45:34 GMT
- Lines: 16
-
-
- The idea of evolving CA rules is an interesting one --- it would not
- be too hard to create a "Blind Watchmaker" style program where the
- user clicks on one of, say, 6 small CA patches to choose to see 6
- mutations of that one, etc. Automated evolution would also be
- reasonable, one might choose a certain entropy level as the fitness
- target. Charles Bennett showed me a rule for the Margolus and Toffoli
- CAM6 which lets the user flip between two active CAs and choose one,
- then see two mutants of that one, etc. I don't know of any other
- existing program that does this kind of thing. There was a paper by
- Norman Packard, I believe, in which he did an automatic evolution of
- a 1D CA towards some goal --- the reference has appeared on this
- newsgroup before. If I get time, I'll try and put an evolving CA
- feature into the "Bugland" program which I am now developing for
- distribution next fall.
-
-