论文信息 - Residual Networks for Computer Go

Residual Networks for Computer Go

Deep learning for the game of Go recently had a tremendous success with the victory of AlphaGo against Lee Sedol in March 2016. We propose to use residual networks so as to improve the training of a policy network for computer Go. Training is faster than with usual convolutional networks and residual networks achieve high accuracy on our test set and a four dan level.

Tristan Cazenave

[1] Serge J. Belongie,et al. Residual Networks Behave Like Ensembles of Relatively Shallow Networks , 2016, NIPS.

[2] Amos J. Storkey,et al. Training Deep Convolutional Neural Networks to Play Go , 2015, ICML.

[3] Yuandong Tian,et al. Better Computer Go Player with Neural Network and Long-term Prediction , 2016, ICLR.

[4] Serge J. Belongie,et al. Residual Networks are Exponential Ensembles of Relatively Shallow Networks , 2016, ArXiv.

[5] Jian Sun,et al. Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[6] Clément Farabet,et al. Torch7: A Matlab-like Environment for Machine Learning , 2011, NIPS 2011.

[7] Demis Hassabis,et al. Mastering the game of Go with deep neural networks and tree search , 2016, Nature.

[8] David Silver,et al. Move Evaluation in Go Using Deep Convolutional Neural Networks , 2014, ICLR.

[9] Rémi Coulom,et al. Efficient Selectivity and Backup Operators in Monte-Carlo Tree Search , 2006, Computers and Games.