Exchangeable Models in Meta Reinforcement Learning
Abstract
One recent approach to meta reinforcement learning (meta-RL) is to integrate models for task inference with models for control. The former component is often based on recurrent neural networks, which do not directly exploit the exchangeable structure of the inputs. We propose to use a lightweight, yet an expressive architecture that accounts for exchangeability. Combined with an off-policy reinforcement learning algorithm, it results in a meta-RL method that is sample-efficient, fast to train and able to quickly adapt to new test tasks as demonstrated on a couple of widely used benchmarks.
Cite
Text
Korshunova et al. "Exchangeable Models in Meta Reinforcement Learning." ICML 2020 Workshops: LifelongML, 2020.Markdown
[Korshunova et al. "Exchangeable Models in Meta Reinforcement Learning." ICML 2020 Workshops: LifelongML, 2020.](https://mlanthology.org/icmlw/2020/korshunova2020icmlw-exchangeable/)BibTeX
@inproceedings{korshunova2020icmlw-exchangeable,
title = {{Exchangeable Models in Meta Reinforcement Learning}},
author = {Korshunova, Iryna and Degrave, Jonas and Dambre, Joni and Gretton, Arthur and Huszar, Ferenc},
booktitle = {ICML 2020 Workshops: LifelongML},
year = {2020},
url = {https://mlanthology.org/icmlw/2020/korshunova2020icmlw-exchangeable/}
}