Open Access

Generalizations of Hölder inequalities for Csiszar’s f-divergence

Journal of Inequalities and Applications20132013:151

https://doi.org/10.1186/1029-242X-2013-151

Received: 12 December 2012

Accepted: 21 March 2013

Published: 3 April 2013

Abstract

In this paper, we establish some new generalizations of the Hölder’s inequality involving Csiszar’s f-divergence of two probability measures. Some related inequalities are also presented.

MSC:26D15, 28A25, 60E15.

Keywords

Csiszar’s discrimination and divergenceHölder inequalitygeneralizationprobabilistic inequality

1 Introduction

Let 1 / p + 1 / q = 1 , assume that f ( x ) and g ( x ) are continuous real-valued functions on [ a , b ] . Then
  1. (1)
    for p > 1 , we have the following Hölder inequality (see [1]):
    a b f ( x ) g ( x ) d x ( a b f p ( x ) d x ) 1 / p ( a b g q ( x ) d x ) 1 / q ;
    (1.1)
     
  2. (2)
    for 0 < p < 1 , we have the following reverse Hölder inequality (see [2]):
    a b f ( x ) g ( x ) d x ( a b f p ( x ) d x ) 1 / p ( a b g q ( x ) d x ) 1 / q .
    (1.2)
     

The above inequalities play an important role in many areas of pure and applied mathematics. A large number of generalizations, refinements, variations and applications of (1.1) and (1.2) have been investigated in the literature (see [311] and references therein). Recently, G.A. Anastassiou [12] established some Hölder’s type inequalities regarding Csiszar’s f-divergence of two probability measures as follows.

Theorem 1.1 (see [12])

Let p , q > 1 such that 1 / p + 1 / q = 1 . Then
Γ | f 1 f 2 | ( μ 1 , μ 2 ) [ Γ | f 1 | p ( μ 1 , μ 2 ) ] 1 / p [ Γ | f 2 | q ( μ 1 , μ 2 ) ] 1 / q .
(1.3)

Theorem 1.2 (see [12])

Let a 1 , a 2 , , a m > 1 , m N , j = 1 m 1 a j = 1 . Then
(1.4)

which is a generalization of Theorem 1.1.

It follows the counterpart of Theorem 1.1.

Theorem 1.3 (see [12])

Let 0 < p < 1 and q < 0 such that 1 / p + 1 / q = 1 , we assume that p 2 > 0 a.e. [ λ ] . Then we have
Γ | f 1 f 2 | ( μ 1 , μ 2 ) [ Γ | f 1 | p ( μ 1 , μ 2 ) ] 1 / p [ Γ | f 2 | q ( μ 1 , μ 2 ) ] 1 / q .
(1.5)

The aim of this paper is to give new generalizations of inequalities (1.4) and (1.5). Some related inequalities are also considered. The paper is organized as follows. In Section 2, we recall some basic facts about the Csiszar’s f-divergence of two probability measures. In Section 3, we will give the main result and its proof.

2 Preliminaries

Assume that f : ( 0 , + ) R is an arbitrary convex function which is strictly convex at 1. As in Csiszar [12, 13], we agree with the following expressions:

Suppose that ( X , A , λ ) is an arbitrary measure space with λ being a finite or σ-finite measure. Let μ 1 , μ 2 be probability measures on X such that μ 1 , μ 2 λ (absolutely continuous).

The Radon-Nikodym derivatives (densities) of μ i with respect to λ is expressed by p i ( x ) :
p i ( x ) = μ i ( d x ) λ ( d x ) , i = 1 , 2 .

Definition 2.1 (see [13])

The f-divergence of the probability measures μ 1 and μ 2 is defined as follows:
Γ f ( μ 1 , μ 2 ) = X p 2 ( x ) f ( p 1 ( x ) p 2 ( x ) ) λ ( d x ) ,

where the function f is named the base function. From Lemma 1.1 of [13], Γ f ( μ 1 , μ 2 ) is always well-defined and Γ f ( μ 1 , μ 2 ) f ( 1 ) with equality only for μ 1 = μ 2 . From [13], we know that Γ f ( μ 1 , μ 2 ) does not depend on the choice of λ. If f ( 1 ) = 0 , then Γ f can be considered as the most general measure of difference between probability measures. For arbitrary convex function f, we notice that Γ f ( μ 1 , μ 2 ) Γ | f | ( μ 1 , μ 2 ) .

The Csiszar’s f-divergence Γ f incorporated most of special cases of probability measure distances, including the variation distance, χ 2 -divergence, information for discrimination or generalized entropy, information gain, mutual information, mean square contingency, etc. Γ f has many applications to almost all applied sciences where stochastics enters. For more references, one can see [1222].

In this paper, we assume that the base function f appearing in the function Γ f have all the above properties of f.

3 Main results

In the section, we establish some new generalizations of the Hölder inequality involving Csiszar’s f-divergence of two probability measures.

Theorem 3.1 Let 0 < a m < 1 , a j < 0 ( j = 1 , 2 , , m 1 ), m N , j = 1 m 1 a j = 1 . Then
(3.1)
Proof Here, we use the generalized Hölder’s inequality (see [23]). We obtain
(3.2)

Hence, we get the desired inequality. □

Theorem 3.2 Let α k j R ( j = 1 , 2 , , m , k = 1 , 2 , , s ), k s 1 a k = 1 , k = 1 s α k j = 0 . Then
  1. (1)
    for a k > 1 , we have the following inequality:
    (3.3)
     
  2. (2)
    for 0 < a s < 1 , a k < 0 ( k = 1 , 2 , , s 1 ), we have the following reverse inequality:
    (3.4)
     
Proof
  1. (1)
    Set
    g k ( x ) = ( j = 1 m f j 1 + a k α k j ( x ) ) 1 / a k .
    (3.5)
     
Applying the assumptions k s 1 a k = 1 and k = 1 s α k j = 0 , we have
k = 1 s g k ( x ) = g 1 g 2 g s = ( j = 1 m f j 1 + a 1 α 1 j ( x ) ) 1 / a 1 ( j = 1 m f j 1 + a 2 α 2 j ( x ) ) 1 / a 2 ( j = 1 m f j 1 + a s α s j ( x ) ) 1 / a s = j = 1 m f j 1 / a 1 + α 1 j ( x ) j = 1 m f j 1 / a 2 + α 2 j ( x ) j = 1 m f j 1 / a s + α s j ( x ) = j = 1 m f j 1 / a 1 + 1 / a 2 + + 1 / a s + α 1 j + α 2 j + + α s j ( x ) = j = 1 m f j ( x ) .
That is,
k = 1 s g k ( x ) = j = 1 m f j ( x ) .
Then we find
(3.6)
By the inequality (1.4), we obtain
(3.7)
In view of (3.5), we have
(3.8)
By (3.6), (3.7) and (3.8), we obtain inequality (3.3).
  1. (2)

    Similar to the proof of inequality (3.3), by (3.5), (3.6), (3.8) and the inequality (3.1), we have inequality (3.4) immediately. □

     
Corrollary 3.1 Under the assumptions of Theorem 3.2, taking s = m , α k j = t / a k for j k and α k k = t ( 1 1 / a k ) with , then we have
  1. (1)
    for α k > 0 , we have the following inequality:
    (3.9)
     
  2. (2)
    for 0 < α m < 1 , α k < 0 ( k = 1 , 2 , , m 1 ), we have the following reverse inequality:
    (3.10)
     
Theorem 3.3 Let ( j = 1 , 2 , , m , k = 1 , 2 , , s ), k s 1 α k = r , k = 1 s α k j = 0 . Then
  1. (1)
    for r α k > 1 , we have the following inequality:
    (3.11)
     
  2. (2)
    for 0 < r α s < 1 , r α k < 0 ( k = 1 , 2 , , s 1 ), we have the following reverse inequality:
    (3.12)
     
Proof (1) Since r α k > 1 and k s 1 α k = r , we get k s 1 r α k = 1 . Then by (3.3), we immediately obtain the inequality (3.11).
  1. (2)

    Since 0 < r α s < 1 , r α k < 0 and k s 1 α k = r , we have k s 1 r α k = 1 , by (3.4), we immediately have the inequality (3.12). This completes the proof. □

     
Theorem 3.4 Under the assumptions of Theorem 3.3, and let s = 2 , α 1 = p , α 2 = q , α 1 j = α 2 j = β j , then
  1. (1)
    for r p > 0 , we have the following inequality:
    (3.13)
     
  2. (2)
    for 0 < r p < 1 , we have the following reverse inequality:
    (3.14)
     
Proof (1) By inequality (1.3), we get
  1. (2)

    Similar to the proof of inequality (3.13), by inequality (1.5), we obtain inequality (3.14). □

     

Remark Assume that X is a finite or countable discrete set, A is its power set P ( X ) and λ has mass 1 for each x X , then Γ f becomes a finite or infinite sum, respectively. As a consequence, all the above obtained integral inequalities are discretized and become summation inequalities.

Declarations

Acknowledgements

Dedicated to Professor Hari M Srivastava.

The authors thank the editor and the referees for their valuable suggestions to improve the quality of this paper.

Authors’ Affiliations

(1)
Department of Computer Engineering, Guangxi Modern Vocational Technology College
(2)
School of Economy & Trade, Zhejiang Industry & Trade Vocational College

References

  1. Mitrinović DS: Analytic Inequalities. Springer, New York; 1970.View ArticleGoogle Scholar
  2. Kuang J: Applied Inequalities. Shandong Science Press, Jinan; 2003.Google Scholar
  3. Hardy G, Littlewood JE, Pólya G: Inequalities. 2nd edition. Cambridge University Press, Cambridge; 1952.Google Scholar
  4. Yang X: A generalization of Hölder inequality. J. Math. Anal. Appl. 2000, 247: 328–330.MathSciNetView ArticleGoogle Scholar
  5. Yang X: Refinement of Hölder inequality and application to Ostrowski inequality. Appl. Math. Comput. 2003, 138: 455–461.MathSciNetView ArticleGoogle Scholar
  6. Yang X: A note on Hölder inequality. Appl. Math. Comput. 2003, 134: 319–322.MathSciNetView ArticleGoogle Scholar
  7. Yang X: Hölder’s inequality. Appl. Math. Lett. 2003, 16: 897–903.MathSciNetView ArticleGoogle Scholar
  8. Wu S, Debnath L: Generalizations of Aczél’s inequality and Popoviciu’s inequality. Indian J. Pure Appl. Math. 2005, 36(2):49–62.MathSciNetGoogle Scholar
  9. He WS: Generalization of a sharp Hölder’s inequality and its application. J. Math. Anal. Appl. 2007, 332: 741–750.MathSciNetView ArticleGoogle Scholar
  10. Wu S: A new sharpened and generalized version of Hölder’s inequality and its applications. Appl. Math. Comput. 2008, 197: 708–714.MathSciNetView ArticleGoogle Scholar
  11. Kwon EG, Bae EK: On a continuous form of Hölder inequality. J. Math. Anal. Appl. 2008, 343: 585–592.MathSciNetView ArticleGoogle Scholar
  12. Anastassiou GA: Hölder-like Csiszar’s type inequalities. Int. J. Pure Appl. Math. 2004, 1: 9–14. www.gbspublisher.comGoogle Scholar
  13. Csiszar I: Information-type measures of difference of probability distributions and indirect observations. Studia Sci. Math. Hung. 1967, 2: 299–318.MathSciNetGoogle Scholar
  14. Anastassiou GA, et al.: Basic optimal approximation of Csiszar’s f -divergence. In Proceedings of 11th Internat. Conf. Approx. Th Edited by: Chui CK. 2004, 15–23.Google Scholar
  15. Anastassiou GA: Fractional and other approximation of Csiszar’s f -divergence. Rend. Circ. Mat. Palermo Suppl. 2005, 76: 197–212.MathSciNetGoogle Scholar
  16. Anastassiou GA: Representations and estimates to Csiszar’s f -divergence. Panam. Math. J. 2006, 16: 83–106.MathSciNetGoogle Scholar
  17. Anastassiou GA: Higher order optimal approximation of Csiszar’s f -divergence. Nonlinear Anal. 2005, 61: 309–339.MathSciNetView ArticleGoogle Scholar
  18. Csiszar I: Eine Informationstheoretische Ungleichung und ihre Anwendung auf den Beweis der Ergodizität von Markoffschen Ketten. Magy. Tud. Akad. Mat. Kut. Intéz. Közl. 1963, 8: 85–108.MathSciNetGoogle Scholar
  19. Csiszar I: On topological properties of f -divergences. Studia Sci. Math. Hung. 1967, 2: 329–339.MathSciNetGoogle Scholar
  20. Dragomir SS: Inequalities for Csiszar f-Divergence in Information Theory. Victoria University, Melbourne; 2000. Edited monograph. On line: http://rgmia.vu.edu.auGoogle Scholar
  21. Anwar M, Hussain S, Pečarić J: Some inequalities for Csiszár-divergence measures. Int. J. Math. Anal. 2009, 3(26):1295–1304.Google Scholar
  22. Kafka P, Österreicher F, Vincze I: On powers of f -divergences defining a distance. Studia Sci. Math. Hung. 1991, 26(4):415–422.Google Scholar
  23. Cheung W-S: Genegralizations of Hölder’s inequality. Int. J. Math. Math. Sci. 2001, 26(1):7–10.MathSciNetView ArticleGoogle Scholar

Copyright

© Chen and Shi; licensee Springer. 2013

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.