dummy-link

ReinforcementLearning

A reinforcement learning package for Julia

Readme

ReinforcementLearning

Documentation Build Status codecov

A reinforcement learning package for Julia.

What is reinforcement learning?

Features

Learning methods

name discrete states linear approximation non-linear approximation
Q-learning/SARSA(λ)
n-step Q-learning/SARSA
Online Policy Gradient
Episodic Reinforce
n-step Actor-Critic Policy-Gradient
Monte Carlo Control
Prioritized Sweeping
(double) DQN

Environments

name state space action space
Cartpole 4D discrete
Mountain Car 2D discrete
Pendulum 3D 1D
Atari pixel images discrete
VizDoom pixel images discrete
POMDPs, MDPs, Mazes, Cliffwalking discrete discrete
OpenAi Gym (using PyCall) see here see here

Preprocessors

  • State Aggregation
  • Tile Coding
  • Random Projections
  • Radial Basis Functions

Helper Functions

  • comparison of different methods
  • callbacks to track performance, change exploration policy, save models during learning etc.

Installation

(v1.0) pkg> add ReinforcementLearning

or in julia v0.6

Pkg.add("ReinforcementLearning")

First Commit

06/07/2018

Last Touched

7 days ago

Commits

72 commits

Used By: