The state-of-art deep rl algorithms for Montezuma's revenge
Montezuma's revenge is the most hardest game to complete in Atari games. So it has been the target of many Deep Reinforcement Learning algorithms. The purpose of this repository is gathering deep-rl algorithms and comparing the performances.
| year / month | the name of algorithm | score of algorithm | use human play data | link |
|---|---|---|---|---|
| 2017 / 10 | RAINBOW | 384 | False | paper code |
| 2016 / 04 | h-DQN | 400 | False | paper |
| 2017 / 03 | Ape-X DQN | 2500 | False | paper |
| 2017 / 03 | FuNs | 2600 | False | paper |
| 2017 / 03 | DQN-PixelCNN | 3700 | False | paper |
| 2017 / 03 | DQFD | 4659 | True | paper |
| 2016 / 06 | DQN-CTS | 6600 | False | paper |
| 2018 / 05 | Ape-X DQFD | 29384 | True | paper |
| 2018 / 05 | TDC+CMC(Youtube) | 41098 | True | paper |
| 2018 / 07 | PPO-reset | 74500 | True | blog code |
| 2018 / 10 | RND | 10000 | False | paper |