SENTINEL: Taming Uncertainty with Ensemble-based Distributional Reinforcement Learning
arXiv:2102.11075
Abstract
In this paper, we consider risk-sensitive sequential decision-making in Reinforcement Learning (RL). Our contributions are two-fold. First, we introduce a novel and coherent quantification of risk, namely composite risk, which quantifies the joint effect of aleatory and epistemic risk during the learning process. Existing works considered either aleatory or epistemic risk individually, or as an additive combination. We prove that the additive formulation is a particular case of the composite risk when the epistemic risk measure is replaced with expectation. Thus, the composite risk is more sensitive to both aleatory and epistemic uncertainty than the individual and additive formulations. We also propose an algorithm, SENTINEL-K, based on ensemble bootstrapping and distributional RL for representing epistemic and aleatory uncertainty respectively. The ensemble of K learners uses Follow The Regularised Leader (FTRL) to aggregate the return distributions and obtain the composite risk. We experimentally verify that SENTINEL-K estimates the return distribution better, and while used with composite risk estimates, demonstrates higher risk-sensitive performance than state-of-the-art risk-sensitive and distributional RL algorithms.
22 pages, 10 figures, 8 tables Accepted for UAI2022
References in corpus (9)
- Robust Adversarial Reinforcement Learning
- Virtual to Real Reinforcement Learning for Autonomous Driving
- Distributional Reinforcement Learning with Quantile Regression
- Decomposition of Uncertainty in Bayesian Deep Learning for Efficient and Risk-sensitive Learning
- Risk-Sensitive and Robust Decision-Making: a CVaR Optimization Approach
- Algorithms for CVaR Optimization in MDPs
- Estimating Risk and Uncertainty in Deep Reinforcement Learning
- Benchmarking Reinforcement Learning Algorithms on Real-World Robots
- Epistemic Risk-Sensitive Reinforcement Learning