This repository contains the code for our paper Control Frequency Adaptation via Action Persistence in Batch Reinforcement Learning, which was accepted at ICML 2020. The FQI implementation is derived from TRLIB.
Python 3
numpy
scikit-learn
joblib
matplotlib
gym
To run PFQI for different persistences in the Cartpole environment:
python3 scripts/run_cartpole.py
The results will be stored in a json file. To plot the results:
python3 plotters/multi_perf_plotter4x4.py plotters/example.json
@incollection{metelli2020control,
author = "Metelli, Alberto Maria and Mazzolini, Flavio and Bisi, Lorenzo and Sabbioni, Luca and Restelli, Marcello",
booktitle = "Proceedings of the 37th International Conference on Machine Learning, Online, PMLR 119, 2020",
pages = "4102--4113",
title = "Control Frequency Adaptation via Action Persistence in Batch Reinforcement Learning",
year = "2020",
}
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent