Skip to content

UCB Exploration via Q-Ensembles

Authors
Richard Y. Chen, Szymon Sidor, Pieter Abbeel, John Schulman
Year
2017
Preprint
Identifiers
arXiv 1706.01502

Abstract

We show how an ensemble of Q^*-functions can be leveraged for more effective exploration in deep reinforcement learning. We build on well established algorithms from the bandit setting, and adapt them to the Q-learning setting. We propose an exploration strategy based on upper-confidence bounds (UCB). Our experiments show significant gains on the Atari benchmark.

Knowledge Base

Map Tree Search

External

Document arXiv Abstract arXiv PDF arXiv HTML Google Scholar OpenAlex

Similar Papers