ML Anthology

Making Efficient Use of Demonstrations to Solve Hard Exploration Problems

Tom Le Paine, Caglar Gulcehre, Bobak Shahriari, Misha Denil, Matt Hoffman, Hubert Soyer, Richard Tanburn, Steven Kapturowski, Neil Rabinowitz, Duncan Williams, Gabriel Barth-Maron, Ziyu Wang, Nando de Freitas, Worlds Team
ICLR 2020
/iclr/2020/paine2020iclr-making/

Abstract

This paper introduces R2D3, an agent that makes efficient use of demonstrations to solve hard exploration problems in partially observable environments with highly variable initial conditions. We also introduce a suite of eight tasks that combine these three properties, and show that R2D3 can solve several of the tasks where other state of the art methods (both with and without demonstrations) fail to see even a single successful trajectory after tens of billions of steps of exploration.

PDF ICLR Semantic Scholar

Cite

Text

Le Paine et al. "Making Efficient Use of Demonstrations to Solve Hard Exploration Problems." International Conference on Learning Representations, 2020.

Markdown

[Le Paine et al. "Making Efficient Use of Demonstrations to Solve Hard Exploration Problems." International Conference on Learning Representations, 2020.](https://mlanthology.org/iclr/2020/paine2020iclr-making/)

BibTeX

@inproceedings{paine2020iclr-making,
  title     = {{Making Efficient Use of Demonstrations to Solve Hard Exploration Problems}},
  author    = {Le Paine, Tom and Gulcehre, Caglar and Shahriari, Bobak and Denil, Misha and Hoffman, Matt and Soyer, Hubert and Tanburn, Richard and Kapturowski, Steven and Rabinowitz, Neil and Williams, Duncan and Barth-Maron, Gabriel and Wang, Ziyu and de Freitas, Nando and Team, Worlds},
  booktitle = {International Conference on Learning Representations},
  year      = {2020},
  url       = {https://mlanthology.org/iclr/2020/paine2020iclr-making/}
}