Skip to content

Reinforcement learning algorithms implemented to learn to play Super Mario Bros.

Notifications You must be signed in to change notification settings

nimishsantosh107/RL-SuperMarioBros

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 

Repository files navigation

RL-SuperMarioBros

Reinforcement learning algorithms implemented to play SuperMario. Will be used as a practical baseline to compare different algorithms & techniques.

Dueling Double DQN

Notes GIFs
21-10-2020
  • epsilon_min (0.01)
  • Very slow learning
  • Score increased rarely post this episode
Episode: 4298
1-1-v0

Dueling Double DQN with Prioritized Replay

Notes GIFs
17-08-2020
  • Algorithm tested
  • Agent didn't learn
NIL
02-10-2020
  • Switched to SmoothL1Loss
  • Too many episodes / Stuck for long (low epsilon_min)
  • Increase epsilon_min (0.01) - should reduce episodes
  • Got unstuck after rare exploration
  • almost beat level
Episode: 4748
1-1-v0
03-10-2020
  • Increased epsilon_min (0.09) used less episodes
  • beat level after more exploration
Episode: 2603
1-4-v0
Episode: 3545
1-4-v0

About

Reinforcement learning algorithms implemented to learn to play Super Mario Bros.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published