Beyond Exponential Utility Functions: A Variance-Adjusted Approach for Risk-Averse Reinforcement Learning
Utility theory has served as a bedrock for modeling risk in economics. Where risk is involved in decision-making, for solving Markov decision processes (MDPs) via utility theory, the exponential utility (EU) function has been used in the literature as an objective function for capturing risk-averse behavior. The EU function framework uses a so-called risk-averseness coefficient (RAC) that seeks to quantify the risk appetite of the decision-maker. Unfortunately, as we show in this paper, the EU framework suffers from computational deficiencies that prevent it from being useful in practice for solution methods based on reinforcement learning (RL). In particular, the value function becomes very large and typically the computer overflows. We provide a simple example to demonstrate this. Further, we show empirically how a variance-adjusted (VA) approach, which approximates the EU function objective for reasonable values of the RAC, can be used in the RL algorithm. The VA framework in a sense has two objectives: maximize expected returns and minimize variance. We conduct empirical studies on a VA-based RL algorithm on the semi-MDP (SMDP), which is a more general version of the MDP. We conclude with a mathematical proof of the boundedness of the iterates in our algorithm.
A. Gosavi et al., "Beyond Exponential Utility Functions: A Variance-Adjusted Approach for Risk-Averse Reinforcement Learning," Proceedings of the IEEE Symposium on Adaptive Dynamic Programming and Reinforcement Learning (2014, Orlando, FL), Institute of Electrical and Electronics Engineers (IEEE), Jan 2014.
The definitive version is available at http://dx.doi.org/10.1109/ADPRL.2014.7010645
IEEE Symposium on Adaptive Dynamic Programming and Reinforcement Learning (2014: Dec. 9-12; Orlando, FL)
Engineering Management and Systems Engineering
Keywords and Phrases
Algorithms; Behavioral research; Computation theory; Decision making; Decision theory; Dynamic programming; Markov processes; Risk analysis; Risks; Computational deficiency; Empirical studies; Exponential utility; Exponential utility function; Markov Decision Processes; Mathematical proof; Objective functions; Reasonable value; Reinforcement learning
International Standard Book Number (ISBN)
International Standard Serial Number (ISSN)
Article - Conference proceedings
© 2014 Institute of Electrical and Electronics Engineers (IEEE), All rights reserved.