A repository of DQN and PPO algorithms implemented in Pytorch Lightning, showcasing a zero-knowledge approach to Tetris.
Modifications made to state representation, reward structure and exploration strategy.
Implementations of VIME-PPO and Bootstrapped DQN.
VIME : https://arxiv.org/abs/1605.09674
Bootstrapped DQN: https://arxiv.org/abs/1602.04621