asifhtamim / reinforcementopenai Goto Github PK

In this project I have tried Reinforcement Learning on two Open AI games . One is Mountain Car Game and another is AirRaid, an Atari Game.

For the Mountain Car game , I have used Q Learning ALgorithm . The Agent needs to earn -110 reward . We have achieved the reward by running upto 30000 episodes. For the simple Q learning algorithm, I have taken help from this blog . The Video of Full training is here . Please notice that the video is full training , so it is very large . Skip to last 1-2 minutes to watch the final Result

For the AirRaid Game, I have used Deep Q Network with Experience replay . For the Architecture I have used the architecture mentioned here .