/rl-montezuma

The state-of-art deep rl algorithms for Montezuma's revenge

rl-montezuma

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.

Algorithms and score on the Montezuma's revenge

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