DocumentCode :
2385103
Title :
On sequential strategies for loss functions with memory
Author :
Merhav, Neri ; Ordentlich, Erik ; Seroussi, Gadiel ; Weinberger, Marcelo J.
Author_Institution :
Dept. of Electr. Eng., Technion-Israel Inst. of Technol., Haifa, Israel
fYear :
2000
fDate :
2000
Firstpage :
70
Abstract :
The sequential decision problem is studied for loss functions with memory and finite action spaces. Based on the theory of Markov decision processes (MDP´s), off-line reference strategies are characterized. An infinite horizon on-line strategy, with corresponding normalized “regret” which is upper-bounded by an O(n-1/3) term for an arbitrary individual sequence of observations of length n, is derived
Keywords :
Markov processes; decision theory; functions; information theory; sequences; Markov decision processes; finite action spaces; infinite horizon on-line strategy; loss functions with memory; normalized regret; off-line reference strategies; sequential decision problem; Costs; Infinite horizon; Laboratories;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Information Theory, 2000. Proceedings. IEEE International Symposium on
Conference_Location :
Sorrento
Print_ISBN :
0-7803-5857-0
Type :
conf
DOI :
10.1109/ISIT.2000.866360
Filename :
866360
Link To Document :
بازگشت