论文信息 - Discretization of (h, phi)-Divergences

Discretization of (h, phi)-Divergences

Abstract In this paper, the problem of loss of information due to the discretization of continuous data and its estimate is studied. It is shown that the rate of convergence of the discretized versions of ( h , ϑ )-divergences to their continuous expressions is quadratic. The results of Ghurye and Johnson (1981) and Zografos, Ferentinos, and Papaioannou (1986) are generalized and supplemented for the ( h , ϑ )-divergence measures.

Miquel Salicrú | María Luisa Menéndez | Takis Papaioannou | K. Ferentinos

[1] K. Ferentinos,et al. Discrete approximations to the Csiszár, Renyi, and Fisher measures of information , 1986 .

[2] J. Burbea. The convexity with respect to Gaussian distributions of divergences of order a , 1984 .

[3] Inder Jeet Taneja,et al. On generalized information and divergence measures and their applications: a brief review , 1989 .

[4] K. Ferentinos,et al. New Parametric Measures of Information , 1981, Inf. Control..

[5] S. G. Ghurye,et al. Discrete approximations to the information integral , 1981 .

[6] R. Serfling. Approximation Theorems of Mathematical Statistics , 1980 .

[7] Inder Jeet Taneja,et al. On Generalized Information Measures and Their Applications , 1989 .

[8] T. Papaioannou,et al. Divergence statistics: sampling properties and multinomial goodness of fit and divergence tests , 1990 .

[9] Calyampudi R. Rao. Diversity and dissimilarity coefficients: A unified approach☆ , 1982 .