I'd be interested to hear, too. Øystein and I made little progress trying to train the nets using TD training, supervised learning against the rollout database, or a combination of both. Any gains we made were marginal, at best.
Øystein did a lot of work on speed improvements, and had some success with that. I don't know if they could easily be integrated back into gnubg, though. -- Ian -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Mark Higgins Sent: 14 January 2013 23:52 To: Philippe Michel Cc: [email protected] Subject: Re: [Bug-gnubg] Is GNUBG actively developed? What training approach have you been using, if you don't mind elaborating? On Jan 14, 2013, at 5:26 PM, Philippe Michel <[email protected]> wrote: > On Fri, 11 Jan 2013, Stelios Togias wrote: > >> I was just wondering if is GnuBG being actively developed. Not so >> much as in the GUI/program part as in the neural networks part. Is it >> considered mature or there's still training and maybe improvement taking >> place? > > I have been training new neural nets (with the input features unchanged) for > some time and I'm almost done with it. > > With them, gnubg should get significantly better on average and (as far as I > can see), suffer from few gross regressions in specific positions. It would > most probably still be a little weaker than eXtreme Gammon 2, though. > > I posted a link to intermediate results there : > http://lists.gnu.org/archive/html/bug-gnubg/2012-09/msg00008.html > > _______________________________________________ > Bug-gnubg mailing list > [email protected] > https://lists.gnu.org/mailman/listinfo/bug-gnubg _______________________________________________ Bug-gnubg mailing list [email protected] https://lists.gnu.org/mailman/listinfo/bug-gnubg _______________________________________________ Bug-gnubg mailing list [email protected] https://lists.gnu.org/mailman/listinfo/bug-gnubg
