Skip to content

Comment on Teaching Deep Convolutional Neural Networks to Play Goparent

Comments

From the paper: " The most obvious next step is to integrate a DCNN into a full fledged Go playing system. For example, a DCNN could be run on a GPU in parallel with a MCTS Go program and be used to provide highly quality priors for what the strongest moves to consider are. Such a system would both be the first to bring sophisticated pattern recognitions abilities to playing Go, and have a strong potential ability to surpass current computer Go programs. "

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.