5 datasets found

A dataset of KGS 0.5-komi games, used to train the value network.

A dataset of KGS 6.5-komi games, used to train the value network.

A dataset of KGS 7.5-komi games, used to train the value network.

A dataset of 30 million self-play games, which with four GTX 980Ti GPUs takes about 3 weeks.

A new approach to a value network architecture for the game Go, called a multi-labelled (ML) value network. The ML value network has the three advantages, offering different...