This is a research project in which I trained neural networks to play the board game mancala using two modern algorithms - deep neuroevolution and deep reinforcement learning. Python and Keras were used to test both algorithms. The deep Q learning algorithm ended up training a neural network with a 98% winrate against a random player.
To read about the full details of the project, you can read the associated research paper via the link below. I compared the efficacy and speed of both algorithms in this paper.