aubreyc/rl_experiments

Some experiements with Reinforcement Learning and Inverse Reinforcement Learning

RL experiments

The repository contains some algorithms of Reinforcement Learning and Inverse Reinforcement Learning.

Installation

Create a virtual env on MacOS:

virtualenv -p /usr/local/bin/python3 virtual_env

Use virtual environment to simplify the setup process:

virtualenv ve_rl_experiments
pip install -r req_virtual_env.txt

Environments

• Gridworld: "Reinforcement Learning: An Introduction" Richard S. Sutton and Andrew G. Barto

Reinforcement learning

• Value Iteration: "Reinforcement Learning: An Introduction" Richard S. Sutton and Andrew G. Barto section 4.4

Inverse Reinforcement Learning:

• Linear Programming IRL - Ng & Russell, 2000 - https://ai.stanford.edu/~ang/papers/icml00-irl.pdf
• Maximum Entropy IRL - Ziebart et al., 2008 - https://www.cs.cmu.edu/~bziebart/publications/maximum-entropy-inverse-reinforcement-learning.pdf

References

Some links to useful material:

• https://github.com/MatthewJA/Inverse-Reinforcement-Learning
• https://github.com/yrlu/irl-imitation