Search found 2 matches

by Daniel Shawul
Sun Jan 31, 2016 7:39 pm
Forum: General Discussion
Topic: AlphaGo
Replies: 4
Views: 14322

Re: AlphaGo

Thanks Alvaro for the nice summary. Their results are indeed mind boggling. When I first heard of CNNs a year or so ago,
I thought they were just yet another way of improving the move selection phase of MCTS...
by Daniel Shawul
Sun Jan 31, 2016 7:34 pm
Forum: MCTS, RL, and neural networks
Topic: MCTS for perft estimation
Replies: 1
Views: 15507

MCTS for perft estimation

Perft (partial game tree size) estimation can be a good testbed for MCTS experiments. Abstract Partial game tree sizes can be estimated using different methods among which the most accurate is conducting monte-carlo simulations of random g...