Reinforcement-Learning

Deep Recurrent Q learning

  • Take successive image as Input of Network
  • By Reccurent Network, Agent can predict Ball moving and move to the direction to block the ball

Structure

  • Image -> CNN -> RNN -> Q value