BYOL-Explore: Exploration by Bootstrapped Prediction
Abstract
We present BYOL-Explore, a conceptually simple yet general approach for curiosity-driven exploration in visually-complex environments. BYOL-Explore learns a world representation, the world dynamics, and an exploration policy all-together by optimizing a single prediction loss in the latent space with no additional auxiliary objective. We show that BYOL-Explore is effective in DM-HARD-8, a challenging partially-observable continuous-action hard-exploration benchmark with visually-rich 3-D environments. On this benchmark, we solve the majority of the tasks purely through augmenting the extrinsic reward with BYOL-Explore s intrinsic reward, whereas prior work could only get off the ground with human demonstrations. As further evidence of the generality of BYOL-Explore, we show that it achieves superhuman performance on the ten hardest exploration games in Atari while having a much simpler design than other competitive agents.
Cite
@article{arxiv.2206.08332,
title = {BYOL-Explore: Exploration by Bootstrapped Prediction},
author = {Zhaohan Daniel Guo and Shantanu Thakoor and Miruna Pîslar and Bernardo Avila Pires and Florent Altché and Corentin Tallec and Alaa Saade and Daniele Calandriello and Jean-Bastien Grill and Yunhao Tang and Michal Valko and Rémi Munos and Mohammad Gheshlaghi Azar and Bilal Piot},
journal= {arXiv preprint arXiv:2206.08332},
year = {2022}
}