UCB exploration via Q-ensembles
OpenAI Blog
June 5, 2017
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.
Verticals
airesearch
Originally published on OpenAI Blog on 6/5/2017