Kenton Born - Researcher, Programmer, Geek!

While taking an artificial intelligence class as an undergraduate, I took a checkers engine made by another person and modified it to make it learn to play better checkers by playing itself thousands of times and “learning from its mistakes”. The engine would play itself over the course of the day, and adjust the weights being used for different aspects of the game (for only one of the computers) according to how the computer did. These weights would then be copied over to the other computer after a few games and the process would repeat itself. Slowly, the engine would find the proper weights of the pieces.

I have attached a paper that discusses it in further detail