I Trained a Deep Q Network Built in TensorFlow to Play Atari Pong

This is simply a transpile &#x2F; reproduction of the original Torch version from Deepmind, but in TensorFlow. It doesn&#x27;t really do anything new or different compared to the paper by [Mnih et. al.](<a href="http:&#x2F;&#x2F;www.nature.com&#x2F;nature&#x2F;journal&#x2F;v518&#x2F;n7540&#x2F;full&#x2F;nature14236.html" rel="nofollow">http:&#x2F;&#x2F;www.nature.com&#x2F;nature&#x2F;journal&#x2F;v518&#x2F;n7540&#x2F;full&#x2F;nature1...</a>) .

Would be interesting to see if it could be regularized to make it a bit less twitchy. (E.g. by giving a fitness bonus to no action &#x2F; and-or penalizing many changes of direction within some time window.)

Github page: <a href="https:&#x2F;&#x2F;github.com&#x2F;asrivat1&#x2F;DeepLearningVideoGames" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;asrivat1&#x2F;DeepLearningVideoGames</a>