AI Game Go Based on Deep Learning

Abstract

With the rapid development of artificial intelligence technology, the game of Go, which is regarded as the peak of human intelligence, has gradually become a new hotspot of AI algorithm research. Based on deep learning technology, this paper selects convolutional neural network and residual network as model architecture, improves the traditional Monte Carlo tree search algorithm, proposes GoMCTS algorithm, introduces strategy value network to guide tree search, and comprehensively utilizes the data generated by human expert chess and self-game. efficient distributed training through parallel computing and parametric server architecture.