[svantana]

Interesting work, however training on data seems unnecessary; chess would be perfect for unsupervised learning - initially it could be trained against an existing chess program, but as the models improve, they could start competing against eachother. Although one would probably need some way of scoring any given board position (compare with DeepMind's Atari playing).

[kylebrown]

If you input a score ("{-1,0,1} on final positions") its effectively a label, that makes the training supervised rather than unsupervised. See [1] for good reasons to be skeptical of unsupervised learning in general.

See [2] for a twist on the DeepMind Atari player. They use Monte Carlo Tree Search (MCTS of automated Go playing fame) to generate training data. By feeding that more carefully generated gameplay data into the deep q-learning net, they exceed DeepMind's (non-MCTS-coupled) performance.

1. http://karpathy.github.io/2014/07/03/feature-learning-escapa...

2. http://www-personal.umich.edu/~rickl/pubs/guo-singh-lee-lewi...

[tmmm]

Chess's too big for this.