DocumentCode :
1145470
Title :
Precise minimax redundancy and regret
Author :
Drmota, Michael ; Szpankowski, Wojciech
Author_Institution :
Inst. fur Diskrete Math. und Geometrie, Tech. Univ. Wien, Vienna, Austria
Volume :
50
Issue :
11
fYear :
2004
Firstpage :
2686
Lastpage :
2707
Abstract :
Recent years have seen a resurgence of interest in redundancy of lossless coding. The redundancy (regret) of universal fixed-to-variable length coding for a class of sources determines by how much the actual code length exceeds the optimal (ideal over the class) code length. In a minimax scenario one finds the best code for the worst source either in the worst case (called also maximal minimax) or on average. We first study the worst case minimax redundancy over a class of stationary ergodic sources and replace Shtarkov´s bound by an exact formula. Among others, we prove that a generalized Shannon code minimizes the worst case redundancy, derive asymptotically its redundancy, and establish some general properties. This allows us to obtain precise redundancy for memoryless, Markov, and renewal sources. For example, we present the exact constant of the redundancy for memoryless and Markov sources by showing that the integer nature of coding contributes log(logm/(m-1))/logm+o(1) where m is the size of the alphabet. Then we deal with the average minimax redundancy and regret. Our approach here is orthogonal to most recent research in this area since we aspire to show that asymptotically the average minimax redundancy is equivalent to the worst case minimax redundancy for some classes of sources. After formulating some general bounds relating these two redundancies, we prove our assertion for memoryless and Markov sources. Nevertheless, we provide evidence that maximal redundancy of renewal processes does not have the same leading term as the average minimax redundancy (however, our general results show that maximal and average regrets are asymptotically equivalent).
Keywords :
Markov processes; maximum likelihood estimation; minimax techniques; redundancy; variable length codes; Markov sources; Shannon code; average-worst case minimax redundancy; exact constant; lossless coding; maximum-likelihood distribution; precise minimax redundancy; stationary ergodic sources; universal fixed-to-variable length coding; Computer science; Entropy; Information theory; Maximum likelihood decoding; Minimax techniques; Source coding; Analytic information theory; average minimax redundancy; generalized Shannon code; maximum-likelihood distribution; minimax and maxmin regrets; sequences $rm mod, 1$; universal modeling; universal noiseless coding; worst case minimax redundancy;
fLanguage :
English
Journal_Title :
Information Theory, IEEE Transactions on
Publisher :
ieee
ISSN :
0018-9448
Type :
jour
DOI :
10.1109/TIT.2004.836702
Filename :
1347356
Link To Document :
بازگشت