Title :
Worst-case bounds for the redundancy of sequential lossless codes and for the logarithmic loss of predictors
Author :
Cesa-Bianchi, Nicolò ; Lugosi, Gábor
Author_Institution :
Polo Didattico e di Ricerca, Milan Univ., Italy
Abstract :
We investigate on-line prediction of individual sequences. Given a class of predictors, the goal is to predict as well as the best predictor in the class, where the loss is measured by the self information (logarithmic) loss function. The excess loss (regret) is closely related to the redundancy of the associated lossless universal code. Using Shtarkov´s theorem (1987) and tools from empirical process theory, we prove a general upper bound on the best possible (minimax) regret. The bound depends on certain metric properties of the class of predictors and is applicable to both parametric and nonparametric classes of predictors
Keywords :
parameter estimation; prediction theory; sequential codes; Shtarkov´s theorem; best predictor; code redundancy; empirical process theory; excess loss; general upper bound; individual sequences; logarithmic loss function; lossless universal code; metric properties; minimax regret; nonparametric predictors; on-line prediction; parametric predictors; predictor logarithmic loss; reference codes; self information; sequential lossless codes; worst-case bounds; Data compression; Economic forecasting; Loss measurement; Maximum likelihood decoding; Maximum likelihood estimation; Minimax techniques; Probability distribution; Time measurement; Upper bound;
Conference_Titel :
Information Theory, 2000. Proceedings. IEEE International Symposium on
Conference_Location :
Sorrento
Print_ISBN :
0-7803-5857-0
DOI :
10.1109/ISIT.2000.866388