Skip to content

gokererdogan/rllib

Repository files navigation

rllib - Reinforcement Learning Library

This library implements various reinforcement learning techniques, including

  • Dynamic programming methods to evaluate policies and find optimal policies
  • A Monte Carlo procedure for evaluating policies
  • Q-learning with lookup tables and function approximators (including neural networks, implemented using theano)
  • Policy gradient for multinomial and normal action distributions with policy functions implemented by neural networks (using theano)

The library also contains an implementation of a game environment (GameEnvironment) for using these methods in two-player games, but these are not tested well at the moment.

About

Reinforcement Learning Library

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages