This is a Bachelor's project from 2018 on applying Reinforcement Learning to the game Lines of Action by Remo Sasso and Quintin van Lohuizen. Temporal difference learning combined with self-play were used resulting in competitive agents. Additionally, look-ahead was implemented and optionally input meta-data can be used.
Requires Java and Deeplearning4j. Run using the file RLonLOA/src/Network.java.
Paper can be found here.