Implementation of AI Poker HULH with Deep CFR and MCCFR ES as thesis for Bachelor of Engineering degree
TeX
PokerBot
Abstract
Main goal of the project is to present author’s implementation of the Deep CFR algorithm with
the Heads Limit Poker Texas Hold’em.
It is a modern method for creating artificial intelligence in large partial-observable games.
Such environments have always been a great challenge and the main barrier
to the development of machine learning. The project will show this problem
by implementation of Deep CFR and analysis of the results.
For this purpose, five recognition models were created every 10 iterations of the algorithm.
Next step was to create games with that models. The results allowed to select the
best playing models and to see how Deep CFR learned over time.
Additionally, the quality of the models was tested against a simple
program which simulate beginner player.