noahgolmant/dqn

Deep Q-Network in tensorflow

DQN

Deep Q-Network in tensorflow.

A Deep Q-Network that learns to play an Atari game through the Atari Learning Environment.

To Train: run python main.py. Currently learns breakout through the breakout.bin ROM.

Requires OpenCV, tensorflow, numpy, and ALE.

dqn's Issues

What does storeExperience ?

Hi, i'm reading your code and everything is clear for me except for

   def storeExperience(self, state, action, reward, newState, terminalState):
        self.replayMemory.append((state, action, reward, newState, terminalState))

I can't figure out why you use it every 100 steps, which purpose it has?