Convolutional Monte Carlo Rollouts in Go
This work addresses the challenge of enhancing AI performance in the domain of Go, an incremental improvement over existing convolutional net-based methods.
The authors tackled the problem of improving Go-playing AI by integrating convolutional networks into all components of a Monte Carlo Tree Search (MCTS) framework, including batch processing, Thompson sampling, and GPU-based rollouts, achieving strong win rates against open-source programs and competitive results against state-of-the-art convolutional net-based Go programs.
In this work, we present a MCTS-based Go-playing program which uses convolutional networks in all parts. Our method performs MCTS in batches, explores the Monte Carlo search tree using Thompson sampling and a convolutional network, and evaluates convnet-based rollouts on the GPU. We achieve strong win rates against open source Go programs and attain competitive results against state of the art convolutional net-based Go-playing programs.