google AlphaGo深度学习算法论文(发布至Nature).pdf

Under review as a conference paper at ICLR 2016 BETTER COMPUTER GO PLAYER WITH NEURAL NET- WORK AND LONG-TERM PREDICTION Yuandong Tian Yan Zhu Facebook AI Research Rutgers University Menlo Park, CA 94025 Facebook AI Research yuandong@ yz328@ ABSTRACT 6 1 0 Competing with top human players in the ancient game of Go has been a long- 2 term goal of artificial intelligence. Go’s high branching factor makes traditional n search techniques ineffective, even on leading-edge hardware, and Go’s evaluation a function could change drastically with one stone change. Recent works [Maddi- J son et al. (2015); Clark Storkey (2015)] show that search is not strictly nec- 6 essary for machine Go players. A pure pattern-matching approach, based on a 2 Deep Convolutional Neural Network (DCNN) that predicts the next move, can perform as well as Monte Carlo Tree Search (MCTS)-based open source Go en- ] G gines such as Pachi [Baudis Gailly (2012)] if its search budget is limited. We extend this idea in our bot named darkforest, which relies on a DCNN designed for L long-term predictions. Darkforest substantially improves the win rate for pattern- . s matching approaches against MCTS-based approaches, even with looser search c