Title :
A universal prediction lemma and applications to universal data compression
Author_Institution :
Dept. of Electr. Eng., Technion-Israel Inst. of Technol., Haifa, Israel
Abstract :
A universal prediction lemma is derived for the class of conditional probability measures that are limited to conditioning events that occur in the training data. The lemma is then used to derive lower bounds on the efficiency of a number of universal data compression algorithms. These bounds are non-asymptotic in the sense that they express the effect of limited training data on the compression efficiency
Keywords :
data compression; prediction theory; probability; conditional probability measures; conditioning events; non-asymptotic bounds; training data; universal data compression; universal prediction lemma; Character generation; Data compression; Electric variables measurement; Entropy; Frequency measurement; Jacobian matrices; Minimax techniques; Statistics; Training data;
Conference_Titel :
Higher-Order Statistics, 1999. Proceedings of the IEEE Signal Processing Workshop on
Conference_Location :
Caesarea
Print_ISBN :
0-7695-0140-0
DOI :
10.1109/HOST.1999.778680
