ISSN:
1573-2878
Keywords:
Partially observed Markov decision problems
;
information quality
;
adaptivity
;
Jensen's inequality
;
Radon-Nikodym derivative
Source:
Springer Online Journal Archives 1860-2000
Topics:
Mathematics
Notes:
Abstract In this paper, it is shown that, if the expected cost-to-go functions generated by a suboptimal design for a partially observed, discrete-time, Markov decision problem with a specific state measurement quality are concave, then the suboptimal design has a desirable adaptivity characteristic relative to that state measurement quality. Optimal strategies are shown to possess this adaptivity characteristic, as does a suboptimal design presented in an example.
Type of Medium:
Electronic Resource
URL:
http://dx.doi.org/10.1007/BF00934845
Permalink