wuyou33 / mta

Python implémentation of "Faster and More Accurate Trace-based Policy Evaluation via Overall Target Error Meta-Optimization"

MTA

Implementation of the paper "Faster and More Accurate Trace-based Policy Evaluation via Overall Target Error Meta-Optimization" [1].

The "ringworld" tests use our implemented version of the environment.

This repository also contains our reproduced $\lambda$-greedy algorithm [2], with some additional tools or scripts to draw the figures showed in the paper [1].

References

[1] Zhao, et al., Faster and More Accurate Trace-based Policy Evaluation via Overall Target Error Meta-Optimization, 2019

[2] White and White, A Greedy Approach to Adapting the Trace Parameter for Temporal Difference Learning, 2016

Requirements

• Python 3.6+
• Dependent python modules

Cite

Please kindly cite our work if necessary:

@article{zhao2019faster,
title={Faster and More Accurate Trace-based Policy Evaluation via Overall Target Error Meta-Optimization},
author={Zhao, Mingde and Porada, Ian and Luan, Sitao and Chang, Xiao-Wen and Precup, Doina},
journal={arXiv},
volume={1904.11439},
year={2019},
url={https://arxiv.org/abs/1904.11439},
}