cornish@RUSSIAN.SPA.SYMBOLICS.COM (jan cornish) (09/26/88)
---- Forwarded Message Follows ---- Return-path: <@AI.AI.MIT.EDU,@WHITE.SWW.Symbolics.COM:cornish@RUSSIAN.SPA.SYMBOLICS.COM> Received: from AI.AI.MIT.EDU by ZERMATT.LCS.MIT.EDU via CHAOS with SMTP id 196188; 22 Sep 88 15:49:58 EDT Received: from WHITE.SWW.Symbolics.COM (TCP 20024234430) by AI.AI.MIT.EDU 22 Sep 88 15:57:01 EDT Received: from RUSSIAN.SPA.Symbolics.COM by WHITE.SWW.Symbolics.COM via CHAOS with CHAOS-MAIL id 209817; Thu 22-Sep-88 12:06:58 PDT Received: from SIERRA-NEVADA.SPA.Symbolics.COM by RUSSIAN.SPA.Symbolics.COM via CHAOS with CHAOS-MAIL id 71086; Tue 20-Sep-88 13:36:49 PDT Date: Tue, 20 Sep 88 13:36 PDT From: jan cornish <cornish@RUSSIAN.SPA.Symbolics.COM> Subject: Genetic Learning Algorithms To: thefool@athena.mit.edu cc: AIList@AI.AI.MIT.EDU, cornish@RUSSIAN.SPA.Symbolics.COM In-Reply-To: The message of 18 Sep 88 06:49 PDT from Michael A. de la Maza <thefool@athena.mit.edu> Message-ID: <19880920203633.9.CORNISH@SIERRA-NEVADA.SPA.Symbolics.COM> Date: 18 Sep 88 13:49:01 GMT From: thefool@athena.mit.edu (Michael A. de la Maza) I am currently working on a genetic learning algorithm(gla) engine that draws inferences from a horse racing database (the results could be enRICHening). Has anyone compiled a bibliography of gla articles/books? If I'm inundated with responses I'll post a summary here. What makes you think a GA will work? You probably would want to use a GA based "classifier system" (see a book by John Holland et. al. called Induction) in which a random population inductive rules are evolved. You might want to take a look at an article "Pinpointing Good Hypothesies with Heuristics" by Steven Salzberg in the book "Artifical Intelligenece & Statistics". He developed a weighted feature vector approach where the weights were updated by heuristics. It worked well. The feature vector was about 70 dimensional. Where are you getting your data? Michael A. de la Maza thefool@athena.mit.edu Query: What is the answer to this question? Answer: what question is this the answer to? [There is a separate list covering genetic algorithms called GA-LIST. Send subscription requests to gref@NRL-AIC.ARPA. However, AIList will continue to carry occasional information ... A good list ... In addition, offutt@caen.engin.umich.edu (Daniel M. Offutt) is offering a GA function optimization package. Contact him for details. - nick]
offutt@CAEN.ENGIN.UMICH.EDU (daniel m offutt) (10/06/88)
> In addition, offutt@caen.engin.umich.edu (Daniel M. Offutt) is > offering a GA function optimization package. Contact him for details. Rather, a package is available from John Grefenstette (gref@nrl-aic.arpa). Ask him for his package called "GENESIS" and ask him to put you on the "GA-LIST" mailing list while you are at it. I have used John's package extensively and can recommend it highly. But I did not write it and it is not available from me. For more details, see my post on genetic algorithms and John's package on sci.physics or sci.math.stat, a few weeks back. ----------------------------------------------------------------------------- Dan Offutt offutt@caen.engin.umich.edu