/Optimal-Options-ICML-2019

Code for generating options for planning and reinforcement learning

Primary LanguagePython

Options for Planning and Reinforcement Learning.

Code for experiments on generating options for planning and reinforcement learning in our 2019 ICML papers:

Jinnai Y. Park JW, Abel D, Konidaris G. 2019. Discovering Options for Exploration by Minimizing Cover Time. Proc. 36th International Conference on Machine Learning.

Jinnai Y, Abel D, Hershkowitz E, Littman M, Konidaris G. 2019. Finding Options that Minimize Planning Time. Proc. 36th International Conference on Machine Learning

Dependencies

The code is written in Python 3. The code is dependent on numpy, scipy, and networkx. To solve MOMI optimally, ortools is required. simple_rl is a library for running RL experiments developed by David Abel. As I made a few tweaks to it, I'm putting the whole code in this repository here.

Directory

graph: approximation algorithms in graph algorithm literature. option_generation: option generation algorithms. experiments: Scripts to replicate experiments in papers.

Example

python3 options/experiments/planning_experiments.py
python3 options/experiments/rl_experiments.py

Author

Yuu Jinnai yuu_jinnai@brown.edu

Simple RL is developed by David Abel.