/rl_from_scratch

Selected Reinforcement Learning algorithms, implemented in Python

Primary LanguagePython

rl_from_scratch

Reinforcement Learning, where agents learn from actively interacting with the environment

Mountain Car learning to reach the flag in a sparse reward environment

About

A selection of Reinforcement Learning Algorithms, mostly tabular methods or environments from OpenAIs gym environment.

Implemented techniques include:

  • Q-Learning
  • Sarsa
  • n-step Sarsa
  • Q(λ)
  • Eligibility Traces
  • Value Iteration

Function approximation / Deep Q-Learning coming soon.

Dependencies

Using Python 3:

pip install -r requirements.txt