Q-Learning on Atari using OpenAI Gym
Project at MPCR for trying new ways of doing RL in an atari environment. This is modified code from https://github.com/tflearn/tflearn/blob/master/examples/reinforcement_learning/atari_1step_qlearning.py
Frame Differences vs. Last Consecutive 4 Frames
Performance on atari games using frame differences was compared to the original from TFLearn examples, which used 4 last states as input to the deep q-network.