Risk-Sensitive and Risk-Neutral Optimality in Markov Decision Chains; a Unified Approach

0377685 - ÚTIA 2013 RIV SK eng C - Conference Paper (international conference)
Sladký, Karel
Risk-Sensitive and Risk-Neutral Optimality in Markov Decision Chains; a Unified Approach.
Quantitative Methods in Economics (Multiple Criteria Decision Making XVI). Bratislava: Vydavatelstvo EKONÓM, 2012 - (Reiff, M.), s. 201-205. ISBN 978-80-225-3426-0.
[Quantitative Methods in Economics (Multiple Criteria Decision Making XVI). Bratislava (SK), 30.05.2012-01.06.2012]
R&D Projects: GA ČR GAP402/11/0150; GA ČR GAP402/10/0956
Institutional support: RVO:67985556
Keywords : discrete-time Markov decision chains * exponential utility functions * risk-sensitive coefficient * connections between risk-sensitive and risk-neutral models
Subject RIV: BB - Applied Statistics, Operational Research
http://library.utia.cas.cz/separaty/2012/E/sladky-risk-sensitive and risk-neutral optimality in markov decision chains a unified approach.pdf

In this note we consider Markov decision chains with finite state space and compact actions spaces where the stream of rewards generated by the Markov processes is evaluated by an exponential utility function (so-called risk-sensitive model) with a given risk sensitivity coefficient. If the risk sensitivity coefficient equals zero (risk-neutral case) we arrive at a standard Markov decision chain. Necessary and sufficient optimality conditions along with equations for average optimal policies both for risk-neutral and risk-sensitive models will be presented and connections and similarity between these approaches will be discussed.
Permanent Link: http://hdl.handle.net/11104/0209781