Skip to main content

Bregman Divergences from Comparative Convexity

  • Conference paper
  • First Online:

Part of the book series: Lecture Notes in Computer Science ((LNIP,volume 10589))

Abstract

Comparative convexity is a generalization of ordinary convexity based on abstract means instead of arithmetic means. We define and study the Bregman divergences with respect to comparative convexity. As an example, we consider the convexity induced by quasi-arithmetic means, report explicit formulas, and show that those Bregman divergences are equivalent to conformal ordinary Bregman divergences on monotone embeddings.

This is a preview of subscription content, log in via an institution.

Buying options

Chapter
USD   29.95
Price excludes VAT (USA)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
eBook
USD   84.99
Price excludes VAT (USA)
  • Available as EPUB and PDF
  • Read on any device
  • Instant download
  • Own it forever
Softcover Book
USD   109.99
Price excludes VAT (USA)
  • Compact, lightweight edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Learn about institutional subscriptions

References

  1. Aczél, J.: A generalization of the notion of convex functions. Det Kongelige Norske Videnskabers Selskabs Forhandlinger, Trondheim 19(24), 87–90 (1947)

    Google Scholar 

  2. Amari, S.: Information Geometry and Its Applications. Applied Mathematical Sciences. Springer, Tokyo (2016)

    Book  MATH  Google Scholar 

  3. Banerjee, A., Merugu, S., Dhillon, I.S., Ghosh, J.: Clustering with bregman divergences. J. Mach. Learn. Res. 6, 1705–1749 (2005)

    MATH  MathSciNet  Google Scholar 

  4. Berrone, L.R., Moro, J.: Lagrangian means. Aequationes Math. 55(3), 217–226 (1998)

    Article  MATH  MathSciNet  Google Scholar 

  5. Boissonnat, J.-D., Nielsen, F., Nock, R.: Bregman Voronoi diagrams. Discrete Comput. Geom. 44(2), 281–307 (2010)

    Article  MATH  MathSciNet  Google Scholar 

  6. Bregman, L.M.: The relaxation method of finding the common point of convex sets and its application to the solution of problems in convex programming. USSR Comput. Math. Math. Phys. 7(3), 200–217 (1967)

    Article  MATH  MathSciNet  Google Scholar 

  7. Bullen, P.S., Mitrinovic, D.S., Vasic, M.: Means and Their Inequalities, vol. 31. Springer, New York (2013)

    MATH  Google Scholar 

  8. Burbea, J., Rao, C.: On the convexity of some divergence measures based on entropy functions. IEEE Trans. Inf. Theory 28(3), 489–495 (1982)

    Article  MATH  MathSciNet  Google Scholar 

  9. De Finetti, B.: Sul concetto di media 3, 369–396 (1931). Istituto italiano degli attuari

    Google Scholar 

  10. Holder, O.L.: Über einen Mittelwertssatz. Nachr. Akad. Wiss. Gottingen Math.-Phys. Kl, pp. 38–47 (1889)

    Google Scholar 

  11. Kolmogorov, A.N.: Sur la notion de moyenne. Acad. Naz. Lincei Mem. Cl. Sci. His. Mat. Natur. Sez. 12, 388–391 (1930)

    Google Scholar 

  12. Matkowski, J.: On weighted extensions of Cauchy’s means. J. Math. Anal. Appl. 319(1), 215–227 (2006)

    Article  MATH  MathSciNet  Google Scholar 

  13. Nagumo, M.: Über eine klasse der mittelwerte. Jpn. J. Math. Trans. Abstracts 7, 71–79 (1930). The Mathematical Society of Japan

    Article  MATH  Google Scholar 

  14. Niculescu, C.P., Persson, L.-E.: Convex Functions and Their Applications: A Contemporary Approach. Springer, New York (2006)

    Book  MATH  Google Scholar 

  15. Nielsen, F., Boltz, S.: The Burbea-Rao and Bhattacharyya centroids. IEEE Trans. Inf. Theory 57(8), 5455–5466 (2011)

    Article  MATH  MathSciNet  Google Scholar 

  16. Nielsen, F., Nock, R.: Total jensen divergences: definition, properties and clustering. In: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 2016–2020. IEEE (2015)

    Google Scholar 

  17. Nock, R., Nielsen, F., Amari, S.: On conformal divergences and their population minimizers. IEEE Trans. Inf. Theory 62(1), 527–538 (2016)

    Article  MATH  MathSciNet  Google Scholar 

  18. Petz, D.: Bregman divergence as relative operator entropy. Acta Mathematica Hung. 116(1–2), 127–131 (2007)

    Article  MATH  MathSciNet  Google Scholar 

  19. Vemuri, B.C., Liu, M., Amari, S., Nielsen, F.: Total bregman divergence and its applications to DTI analysis. IEEE Trans. Med. Imaging 30(2), 475–483 (2011)

    Article  Google Scholar 

  20. Zhang, J.: Divergence function, duality, and convex analysis. Neural Comput. 16(1), 159–195 (2004)

    Article  MATH  Google Scholar 

  21. Zhang, J.: Nonparametric information geometry: From divergence function to referential-representational biduality on statistical manifolds. Entropy 15(12), 5384–5418 (2013)

    Article  MATH  MathSciNet  Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Frank Nielsen .

Editor information

Editors and Affiliations

A Quasi-arithmetic to Ordinary Convexity Criterion

A Quasi-arithmetic to Ordinary Convexity Criterion

Lemma 1

( \((\rho ,\tau )\) -convexity \(\leftrightarrow \) ordinary convexity [1]). Let \(\rho :I\rightarrow \mathbb {R}\) and \(\tau :J\rightarrow \mathbb {R}\) be two continuous and strictly monotone real-valued functions with \(\tau \) increasing, then function \(F:I\rightarrow J\) is \((\rho ,\tau )\)-convex iff function \(G=F_{\rho ,\tau } = \tau \circ F\circ \rho ^{-1}\) is (ordinary) convex on \(\rho (I)\).

Proof

Let us rewrite the \((\rho ,\tau )\)-convexity midpoint inequality as follows:

$$\begin{aligned} F(M_\rho (x,y))\le & {} M_\tau (F(x),F(y)),\\ F\left( \rho ^{-1}\left( \frac{\rho (x)+\rho (y)}{2}\right) \right)\le & {} \tau ^{-1}\left( \frac{\tau (F(x))+\tau (F(y))}{2}\right) , \end{aligned}$$

Since \(\tau \) is strictly increasing, we have:

$$\begin{aligned} (\tau \circ F\circ \rho ^{-1}) \left( \frac{\rho (x)+\rho (y)}{2}\right) \le \frac{(\tau \circ F)(x)+(\tau \circ F)(y)}{2}. \end{aligned}$$
(17)

Let \(u=\rho (x)\) and \(v=\rho (y)\) so that \(x=\rho ^{-1}(u)\) and \(y=\rho ^{-1}(v)\) (with \(u,v\in \rho (I)\)). Then it comes that:

$$\begin{aligned} (\tau \circ F\circ \rho ^{-1})\left( \frac{u+v}{2}\right) \le \frac{(\tau \circ F\circ \rho ^{-1})(u)+(\tau \circ F\circ \rho ^{-1})(v)}{2}. \end{aligned}$$
(18)

This last inequality is precisely the ordinary midpoint convexity inequality for function \(G=F_{\rho ,\tau }=\tau \circ F\circ \rho ^{-1}\). Thus a function F is \((\rho ,\tau )\)-convex iff \(G=\tau \circ F\circ \rho ^{-1}\) is ordinary convex, and vice-versa.

Rights and permissions

Reprints and permissions

Copyright information

© 2017 Springer International Publishing AG

About this paper

Cite this paper

Nielsen, F., Nock, R. (2017). Bregman Divergences from Comparative Convexity. In: Nielsen, F., Barbaresco, F. (eds) Geometric Science of Information. GSI 2017. Lecture Notes in Computer Science(), vol 10589. Springer, Cham. https://doi.org/10.1007/978-3-319-68445-1_74

Download citation

  • DOI: https://doi.org/10.1007/978-3-319-68445-1_74

  • Published:

  • Publisher Name: Springer, Cham

  • Print ISBN: 978-3-319-68444-4

  • Online ISBN: 978-3-319-68445-1

  • eBook Packages: Computer ScienceComputer Science (R0)

Publish with us

Policies and ethics