Christopher Clark

From Chessprogramming wiki
Revision as of 11:23, 16 November 2019 by GerdIsenberg (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Home * People * Christopher Clark

Christopher Clark [1]

Christopher Clark,
an American computer scientist and researcher at Allen Institute for Artificial Intelligence, Seattle, Washington, and Ph.D. student at the University of Washington. He received a M.Sc. in artificial intelligence from the University of Edinburgh, where he studied machine learning and completed a thesis on using deep convolutional neural networks for the game of Go to learn and predict the moves made by expert Go players.

DCNNs in Go

As reported in their 2014 paper Teaching Deep Convolutional Neural Networks to Play Go, Clark and Amos Storkey trained an 8-layer convolutional neural network [2] by supervised learning from a database of human professional games to predict the moves made by expert Go players. They introduced a number of novel techniques, including a method of tying weights in the network to 'hard code' symmetries that are expect to exist in the target function, and demonstrated in an ablation study they considerably improve performance. Their final networks can consistently defeat Gnu Go, indicating it is state of the art among programs that do not use Monte-Carlo Tree Search, and was also able to win some games against Fuego while using a fraction of the playing time [3] [4] [5] [6] [7]

Selected Publications


External Links


Up one level