In this version, I have used Dr. Norvig's data structures, but wrapped it so it can display on an HTML canvas, and a person can play against it on the canvas.
Some of the stuff going on here is:
- Alpha-Beta game tree searching
- Position evaluation using features from Logistello
- Learning via Samuel's strategy of using the look-ahead to train
the evaluation function
- Using a Naive Bayes Classifier instead of Logistic Regression
- Some HTML5 stuff, like application caching, and maybe a webworker in the future.