Hi Petr > I guess so. If you want to also to move up to 19x19, try adding > priors to your moves based on the Common Fate Graph distance to the last > move; I think your program should get to at least 5k KGS at that point > if you have everything well debugged.
CFG's is on my list of things to add. Ideally, I want to add it as another ELO feature. > Please also note that the original paper describes a program that does > not contain RAVE; many plain UCT improvements are rendered ineffective > and overshadown wrt. RAVE (which is such a - disappointingly, for > further research - superb hauristic). Some people managed to get it work > with RAVE too. Another possible suspect might be progressive unpruning. I didn't think of that. I will have a look at the difference without RAVE. I was planning on adding progressive unpruning next, but I was hoping for an improvement from ELO features in playouts first. Guess you don't always get what you hope for :) -- Francois van Niekerk Email: [email protected] | Twitter: @francoisvn Cell: +2784 0350 214 | Website: http://leafcloud.com On Fri, Dec 31, 2010 at 3:01 AM, Petr Baudis <[email protected]> wrote: > Hi! > > On Thu, Dec 30, 2010 at 07:08:27PM +0200, Francois van Niekerk wrote: >> Some details about Oakfoam: >> - UCT algo (surprise, surprise ;)) >> - RAVE >> - Mogo 3x3 patterns >> - Open Source under the BSD license >> - Almost everything is adjustable at runtime using parameters >> - Achieved a 1700 ELO rating on CGOS 9x9 recently >> - Repo at http://bitbucket.org/francoisvn/oakfoam/ > > Awesome! Best of luck to you. I think some people have been looking > for a BSD-licenced engine with reasonable playing strength, so this will > make them happy too. :-) > >> I have been working almost exclusively on 9x9. I would also like to >> mention that most of my parameters have not been tuned, so when I get >> around to that I should get some more "free" strength. However, my >> program mostly seems to be comparable to others when using UCT+RAVE so >> I'm satisfied for now. For reference I need about 100k playouts with >> RAVE to get 50% winrate against GnuGo 3.8 L10. Does this seem in >> order? > > I guess so. If you want to also to move up to 19x19, try adding > priors to your moves based on the Common Fate Graph distance to the last > move; I think your program should get to at least 5k KGS at that point > if you have everything well debugged. > >> So my questions are: Does anyone know where I might have gone wrong? >> Is there a way for me to better verify that my feature gammas are ok? > > I have also tried to adopt the Elo way and failed, with similar > results to yours, and I know others that did. And I also know some > people that managed to get it working beautifully. So there is a trick > somewhere along the way and it is unfortunately not known widely. > > Please also note that the original paper describes a program that does > not contain RAVE; many plain UCT improvements are rendered ineffective > and overshadown wrt. RAVE (which is such a - disappointingly, for > further research - superb hauristic). Some people managed to get it work > with RAVE too. Another possible suspect might be progressive unpruning. > > -- > Petr "Pasky" Baudis > Computer science education cannot make an expert programmer any more > than studying brushes and pigment can make an expert painter. --esr > _______________________________________________ > Computer-go mailing list > [email protected] > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go > _______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
