Geometry Induced by a Generalization of Renyi Divergence

被引:13
作者
de Souza, David C. [1 ]
Vigelis, Rui F. [2 ]
Cavalcante, Charles C. [3 ]
机构
[1] Inst Fed Ceara, Campus Maracanau, BR-61939140 Fortaleza, Ceara, Brazil
[2] Univ Fed Ceara, Comp Engn Sch, Campus Sobral, BR-62010560 Sobral, Brazil
[3] Univ Fed Ceara, Dept Teleinformat Engn, BR-60455900 Fortaleza, Ceara, Brazil
来源
ENTROPY | 2016年 / 18卷 / 11期
关键词
Renyi divergence; phi-function; phi-divergence; phi-family; statistical manifold; information geometry; DEFORMED EXPONENTIAL-FAMILIES; SPACE;
D O I
10.3390/e18110407
中图分类号
O4 [物理学];
学科分类号
0702 ;
摘要
In this paper, we propose a generalization of Renyi divergence, and then we investigate its induced geometry. This generalization is given in terms of a phi-function, the same function that is used in the definition of non-parametric phi-families. The properties of phi-functions proved to be crucial in the generalization of Renyi divergence. Assuming appropriate conditions, we verify that the generalized Renyi divergence reduces, in a limiting case, to the phi-divergence. In generalized statistical manifold, the phi-divergence induces a pair of dual connections relation D (a) = 1 a 2 D (1) + 1 + a 2 D (1), with alpha is an element of [-1, 1].
引用
收藏
页数:16
相关论文
共 33 条
  • [1] Amari S.I., 2000, Methods of Information Geometry, DOI DOI 10.1090/MMONO/191
  • [2] Geometry of deformed exponential families: Invariant, dually-flat and conformal geometries
    Amari, Shun-ichi
    Ohara, Atsumi
    Matsuzoe, Hiroshi
    [J]. PHYSICA A-STATISTICAL MECHANICS AND ITS APPLICATIONS, 2012, 391 (18) : 4308 - 4319
  • [3] Geometry of q-Exponential Family of Probability Distributions
    Amari, Shun-ichi
    Ohara, Atsumi
    [J]. ENTROPY, 2011, 13 (06): : 1170 - 1185
  • [5] [Anonymous], LECT NOTES COMPUT SC
  • [6] [Anonymous], [No title captured]
  • [7] [Anonymous], 1945, BULL CALCUTTA MATH S, DOI DOI 10.1007/978-1-4612-0919-5_16
  • [8] [Anonymous], 1985, DIFFERENTIAL GEOMETR
  • [9] [Anonymous], 2007, GEOMETRY HESSIAN STR, DOI DOI 10.1142/6241
  • [10] Banerjee A, 2005, J MACH LEARN RES, V6, P1705