Optimal solution for undiscounted variance penalized Markov decision chains. Abstract

0410871 - UTIA-B 20020085 DE eng A - Abstrakt
Sladký, Karel
Optimal solution for undiscounted variance penalized Markov decision chains. Abstract.
Berlin: HumboldtUniversity Berlin, 2002. Mathematical Methods in Economy and Industry. Abstracts. s. 14
[Joint Czech-German-Slovak Conference /12./. 22.07.2002-26.07.2002, Arnstadt]
Grant CEP: GA ČR GA402/02/1015; GA ČR GA402/02/0539
Výzkumný záměr: CEZ:AV0Z1075907
Klíčová slova: Markov decision chains * optimal policies * mean-variance penalization
Kód oboru RIV: BB - Aplikovaná statistika, operační výzkum

We investigate how the mean variance selection rule can work in Markovian decision models. In contrast to the classical models we assume that instead of maximizing the mean reward per transition we consider more sophisticated criteria taking into account also higher moments and the variance of the cumulative reward. Properties of optimal policies as well as optimization procedures with respect to the above criteria are discussed primarily for undiscounted long run models.
Trvalý link: http://hdl.handle.net/11104/0130958