Epistemic risk-sensitive reinforcement learning
Paper in proceeding, 2020

We develop a framework for risk-sensitive behaviour in reinforcement learning (RL) due to uncertainty about the environment dynamics by leveraging utility-based definitions of risk sensitivity. In this framework, the preference for risk can be tuned by varying the utility function, for which we develop dynamic programming (DP) and policy gradient-based algorithms. The risk-averse behavior is compared with the behavior of risk-neutral policy in environments with epistemic risk.

Author

Hannes Eriksson

Chalmers, Computer Science and Engineering (Chalmers), Data Science

Zenuity AB

Christos Dimitrakakis

University of Oslo

Zenuity AB

ESANN 2020 - Proceedings, 28th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning

339-344

28th European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, ESANN 2020
Virtual, Online, Belgium,

Subject Categories

Philosophy

Transport Systems and Logistics

Other Civil Engineering

More information

Latest update

4/28/2021