The Potential of the Return Distribution for Exploration in RL

2018-06-11Code Available0· sign in to hype

Thomas M. Moerland, Joost Broekens, Catholijn M. Jonker

Code Available — Be the first to reproduce this paper.

Code

github.com/tmoer/return_distribution_exploration
OfficialIn papertf★ 0

Abstract

This paper studies the potential of the return distribution for exploration in deterministic reinforcement learning (RL) environments. We study network losses and propagation mechanisms for Gaussian, Categorical and Gaussian mixture distributions. Combined with exploration policies that leverage this return distribution, we solve, for example, a randomized Chain task of length 100, which has not been reported before when learning with neural networks.

Tasks

reinforcement-learning Reinforcement Learning Reinforcement Learning (RL)

The Potential of the Return Distribution for Exploration in RL

Code

Abstract

Tasks

Reproductions