Skip to content

taoketao/Path-Finder

Repository files navigation

Path-Finder

Implementation of a deep reinforcement learning model trained to solve pathfinder puzzles.

See results commit on 4-22-17. Observation: convergence and convergence rate are relatively unphased by maximum number of actions except when the number is insufficient.

5/28 todo

  • implement scheduler:
    1. learning rate scheduler (1.5. MNA scheduler)
    2. state presentation scheduler
  • implement deuling architectures (easy modification)

About

Implementation of a deep reinforcement learning model trained to solve pathfinder puzzles

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors