This research aims to explore the field of Multi-Task Reinforcement Learning. With a slight modification to the Deep Q-Network algorithm, Polygame AI is capable of learning to play several Atari games. Future work involves scaling our algorithm up to a more complex and diverse set of tasks.
Utilizes CleanRL's dqn-atari code as a base, found in this git repository: https://github.com/vwxyzjn/cleanrl/blob/master/LICENSE