Saturday, November 07, 2015

[hhhalctv] Learning where humans fail

It seems a relatively straightforward application of machine learning to learn in which kinds of chess positions humans tend to make mistakes: use a database of human games and computer evaluation of each position.  One could also apply ML to a particular human, either opponent modeling or improving oneself.

Can this be generalized?  We need other large training sets of examples of humans being correct or making mistakes.

No comments :