Re:Record-breaking win in computer Go match vs professional ...
by tansey on Monday March 19, 2012 @12:43AM
(#3722350)
Is there a paper behind the design of Zen? I'm assuming it's some combination of heuristics and MCTS.
[ Parent
| on HN ]
Re:Record-breaking win in computer Go match vs professional ...
by mquander on Monday March 19, 2012 @01:05AM
(#3722403)
I don't know of a paper; it's sold commercially by the author and is closed-source. It's definitely fair to say that it is a combination of MCTS and heuristics. The best I can find at a glance is this post on the computer-go list, where the author remarks:
Zen is similar to partly MoGo, partly Crazy Stone. It has shape patterns that are generated by minorization-maximization algorithm like Crazy Stone. But they are directly combined with UCT - I don't use progressive widening. Probably the most original part of Zen is in the playout. I don't think MC simulations must be always fast, so it has a lot of hard-coded Go knowledge.