TY - JOUR
T1 - Ergodicity, decisions, and partial information
AU - van Handel, Ramon
N1 - Funding Information:
This work was partially supported by NSF grant DMS-1005575.
Publisher Copyright:
© Springer International Publishing Switzerland 2014.
PY - 2014
Y1 - 2014
N2 - In the simplest sequential decision problem for an ergodic stochastic process X, at each time n a decision un is made as a function of past observations X0; : : : ;Xn_1, and a loss l.un;Xn/ is incurred. In this setting, it is known that one may choose (under a mild integrability assumption) a decision strategy whose pathwise time-average loss is asymptotically smaller than that of any other strategy. The corresponding problem in the case of partial information proves to be much more delicate, however: if the process X is not observable, but decisions must be based on the observation of a different process Y , the existence of pathwise optimal strategies is not guaranteed. The aim of this paper is to exhibit connections between pathwise optimal strategies and notions from ergodic theory. The sequential decision problem is developed in the general setting of an ergodic dynamical system .˝;B; P; T / with partial information Y _ B. The existence of pathwise optimal strategies grounded in two basic properties: the conditional ergodic theory of the dynamical system, and the complexity of the loss function. When the loss function is not too complex, a general sufficient condition for the existence of pathwise optimal strategies is that the dynamical system is a conditional K-automorphism relative to the past observationsWn_0 T nY. If the conditional ergodicity assumption is strengthened, the complexity assumption can be weakened. Several examples demonstrate the interplay between complexity and ergodicity, which does not arise in the case of full information. Our results also yield a decision-theoretic characterization of weak mixing in ergodic theory, and establish pathwise optimality of ergodic nonlinear filters.
AB - In the simplest sequential decision problem for an ergodic stochastic process X, at each time n a decision un is made as a function of past observations X0; : : : ;Xn_1, and a loss l.un;Xn/ is incurred. In this setting, it is known that one may choose (under a mild integrability assumption) a decision strategy whose pathwise time-average loss is asymptotically smaller than that of any other strategy. The corresponding problem in the case of partial information proves to be much more delicate, however: if the process X is not observable, but decisions must be based on the observation of a different process Y , the existence of pathwise optimal strategies is not guaranteed. The aim of this paper is to exhibit connections between pathwise optimal strategies and notions from ergodic theory. The sequential decision problem is developed in the general setting of an ergodic dynamical system .˝;B; P; T / with partial information Y _ B. The existence of pathwise optimal strategies grounded in two basic properties: the conditional ergodic theory of the dynamical system, and the complexity of the loss function. When the loss function is not too complex, a general sufficient condition for the existence of pathwise optimal strategies is that the dynamical system is a conditional K-automorphism relative to the past observationsWn_0 T nY. If the conditional ergodicity assumption is strengthened, the complexity assumption can be weakened. Several examples demonstrate the interplay between complexity and ergodicity, which does not arise in the case of full information. Our results also yield a decision-theoretic characterization of weak mixing in ergodic theory, and establish pathwise optimality of ergodic nonlinear filters.
UR - http://www.scopus.com/inward/record.url?scp=84927131880&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84927131880&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-11970-0_18
DO - 10.1007/978-3-319-11970-0_18
M3 - Article
AN - SCOPUS:84927131880
VL - 2123
SP - 411
EP - 459
JO - Lecture Notes in Mathematics
JF - Lecture Notes in Mathematics
SN - 0075-8434
ER -