摘要: We study the problem of data compression, gambling and prediction a sequence x/sup n/ = x/sub 1/x/sub 2/...x/sub from certain alphabet X, in terms regret (Shtarkov 1988) redundancy with respect to general exponential family, smooth also Markov sources. In particular, we show that variants Jeffreys mixture asymptotically achieve their minimax values.
elsevier.com
yale.edu 
sci-hub.se 参考文章(6)
Bertrand S Clarke, Andrew R Barron, None, Jeffreys' prior is asymptotically least favorable under entropy risk Journal of Statistical Planning and Inference. ,vol. 41, pp. 37- 60 ,(1994) , 10.1016/0378-3758(94)90153-8
J.J. Rissanen, Fisher information and stochastic complexity IEEE Transactions on Information Theory. ,vol. 42, pp. 40- 47 ,(1996) , 10.1109/18.481776
Qun Xie, Andrew R Barron, None, Minimax redundancy for the class of memoryless sources IEEE Transactions on Information Theory. ,vol. 43, pp. 646- 657 ,(1997) , 10.1109/18.556120
Qun Xie, A.R. Barron, Asymptotic minimax regret for data compression, gambling and prediction international symposium on information theory. ,vol. 46, pp. 431- 445 ,(1997) , 10.1109/18.825803
L. Davisson, A. Leon-Garcia, A source matching approach to finding minimax codes IEEE Transactions on Information Theory. ,vol. 26, pp. 166- 174 ,(1980) , 10.1109/TIT.1980.1056167
Yu. M. Shtarkov, AIM FUNCTIONS AND SEQUENTIAL ESTIMATION OF THE SOURCE MODEL FOR UNIVERSAL CODING Problems of Information Transmission. ,vol. 35, pp. 275- 285 ,(1999)