/DQN

Play the breakout with DQN. Fork from https://github.com/zhengsizuo/DRL_udacity.git. Reference to https://github.com/udacity/deep-reinforcement-learning/tree/master/dqn

Primary LanguagePython

Implementation

  • q_model.py:build the same CNN architecture as in the original Nature paper
  • dqn_agent.py:implentation of replay buffer and target network techniques
  • Deep_Q_Network.py:train the Atari game Breakout

Results

Result of training for 30 hours using GTX TITAN.

best

30 thousand iterations