On Wed, Nov 12, 2008 at 11:27 AM, Christian Anthon <[EMAIL PROTECTED]> wrote: > Hi Joseph, > > I'll gladly take a look at your dobling code, can you point me to the > location in the code? > > What I meant was that it was partly a fault in the neural net, I > didn't assume you had written the resign code in gnubg.
OK. I would not have been surprised at the rejection if I was playing at ply 0. I was surprised only because I set play evaluation at 2 ply. You can work your way up from gnubg/gnubg-nn/analyze/danalyze.cc Personally, I would start with http://fibs2html.sourceforge.net/ code. -Joseph > > Christian. > > On Tue, Nov 11, 2008 at 9:59 PM, Joseph Heled <[EMAIL PROTECTED]> wrote: >> On Tue, Nov 11, 2008 at 10:14 PM, Christian Anthon >> <[EMAIL PROTECTED]> wrote: >>> Well Joseph you are partly to blame :). >> >> Well, that shows you had not been around as long as I am :) I did not >> write this code. My code always used a different (and I think >> superior) cube handling method, and always worked at any ply. it is in >> the gnubg-nn branch if you are interested. it is C++. >> >> -Joseph >> >>> Accept/Reject decision were >>> investigated at 0ply regardless of the player settings, and 0ply >>> believes that you loose 1% gammon here. I'm going to up the test to >>> 2ply, it won't be perfect, but certainly better. Alternatively we >>> could run a one-sided rollout, if somebody thinks it is necessary? >>> >>> Christian. >>> >>> On Sun, Nov 9, 2008 at 3:49 AM, Joseph Heled <[EMAIL PROTECTED]> wrote: >>>> I offer to resign normal. gnubg declines. why? >>>> >>>> KwAA4L61BAAAAA:UQngAUAAUAAA >>>> >>>> >>>> _______________________________________________ >>>> Bug-gnubg mailing list >>>> [email protected] >>>> http://lists.gnu.org/mailman/listinfo/bug-gnubg >>>> >>> >> > _______________________________________________ Bug-gnubg mailing list [email protected] http://lists.gnu.org/mailman/listinfo/bug-gnubg
