Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller
We present the first deep learning model to successfully learn control policies directly from high-dimensional sensory input using reinforcement learning. The model is a convolutional neural network, trained with a variant of Q-learning, whose input is raw pixels and whose output is a value function estimating future rewards. We apply our method to seven Atari 2600 games from the Arcade Learning Environment, with no adjustment of the architecture or learning algorithm. We find that it outperforms all previous approaches on six of the games and surpasses a human expert on three of them.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Atari Games | Atari 2600 Pong | Score | 21 | DQN Best |
| Atari Games | Atari 2600 Enduro | Score | 661 | DQN Best |
| Atari Games | Atari 2600 Breakout | Score | 225 | DQN Best |
| Atari Games | Atari 2600 Space Invaders | Score | 1075 | DQN Best |
| Atari Games | Atari 2600 Beam Rider | Score | 5184 | DQN Best |
| Atari Games | Atari 2600 Seaquest | Score | 1740 | DQN Best |
| Atari Games | Atari 2600 Q*Bert | Score | 4500 | DQN Best |
| Video Games | Atari 2600 Pong | Score | 21 | DQN Best |
| Video Games | Atari 2600 Enduro | Score | 661 | DQN Best |
| Video Games | Atari 2600 Breakout | Score | 225 | DQN Best |
| Video Games | Atari 2600 Space Invaders | Score | 1075 | DQN Best |
| Video Games | Atari 2600 Beam Rider | Score | 5184 | DQN Best |
| Video Games | Atari 2600 Seaquest | Score | 1740 | DQN Best |
| Video Games | Atari 2600 Q*Bert | Score | 4500 | DQN Best |