TensorFlow implementation of "Trajectory-wise Multiple Choice Learning for Dynamics Generalization in Reinforcement Learning" (NeurIPS 2020).
Trajectory-wise multiple choice learning (T-MCL) learns a multi-headed dynamics model for dynamics generalization. To effectively utilize specialized prediction heads, prediction heads are adaptively selected at evaluation time.
Install required packages with below commands:
conda create -n tmcl python=3.6
pip install -r requirements.txt
Train and evaluate agents:
python -m run_scripts.run_tmcl --dataset [hopper/slim_humanoid/halfcheetah/cripple_ant] --normalize_flag
@inproceedings{seo2020trajectory,
title={Trajectory-wise Multiple Choice Learning for Dynamics Generalization in Reinforcement Learning},
author={Seo, Younggyo and Lee, Kimin and Clavera, Ignasi and Kurutach, Thanard and Shin, Jinwoo and Abbeel, Pieter},
booktitle={Advances in Neural Information Processing Systems},
year={2020}
}