Implemented Deep Q Learning agent to play Atari games provided by OpenAI gym in Pytorch Improved training efficiency of the model by implementing techniques like Experience Replay and Fixed Q targets Used other architectures like Double DQN, Dueling DQN and Dueling Double DQN to shorten learning time Achieved 100% win rate in just 102 games.
episode: 103 score: 3.0 average score -15.0 best score -15.25 epsilon 0.10 steps 178231