AI bot for checkers based on AlphaZero
Summary
With the growing domain of AI solutions we always must stay in touch with the state of the art research. This is why one of our fields is Reinforcement Learning, a young but promising study. We have created a checkers bot solely with the power of RL as a result of our pursuit and drive to learn and apply this methodology to business needs.
Project duration:
13 months
Team
3 people
3 Researchers
Technologies
Python, C++, Reinforcement Learning
Tech challenge
From scratch implementation of some of the most complex algorithms in history.
High optimization of every function.
Multi-GPU model training.
Solution
Implemented Monte Carlo Tree Search. Implemented complete self-play framework from scratch. Implemented on bits operations highly optimized game environment. It won every person on our team.
