Skip to content

We use the Rainbow DQN model to build agents that play Ms-Pacman, Atlantis and Demon Attack. We make modifications to the model that allow much faster convergence on Ms-Pacman with respect to Deepmind's original paper and obtain comparable performance.

Notifications You must be signed in to change notification settings

willi-menapace/atari_reinforcement_learning

Repository files navigation

Rainbow DQN on Atari games

We use the Rainbow DQN model to build agents that play Ms-Pacman, Atlantis and Demon Attack. We make modifications to the model that allow much faster convergence on Ms-Pacman with respect to Deepmind's original paper and obtain comparable performance.

Read our results

Thanks to Max Lapan for his useful starting codebase.

About

We use the Rainbow DQN model to build agents that play Ms-Pacman, Atlantis and Demon Attack. We make modifications to the model that allow much faster convergence on Ms-Pacman with respect to Deepmind's original paper and obtain comparable performance.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages