Implementation of the paper "META: State-based Eligibility Traces for More Sample-Efficient Policy Evaluation" [1]. Watch the video of the paper here: https://www.youtube.com/watch?v=3Ud8Ils1_mo

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

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

[1] Zhao et al., META-Learning State-based Eligibility Traces for More Sample-Efficient Policy Evaluation, 2019

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

• Python 3.6+
• Numpy, Numba
• OpenAI Gym
• Dependent python modules

Please kindly cite our work if necessary:

@inproceedings{zhao2020meta,
title={META: State-based Eligibility Traces for More Sample-Efficient Policy Evaluation},
author={Zhao, Mingde and Luan, Sitao and Porada, Ian and Chang, Xiao-Wen and Precup, Doina},
booktitle = {International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS)},
year = {2020},
url={https://arxiv.org/abs/1904.11439}
}