On fine properties of mixtures with respect to concentration of measure and Sobolev type inequalities
Annales de l'I.H.P. Probabilités et statistiques, Volume 46 (2010) no. 1, p. 72-96

Mixtures are convex combinations of laws. Despite this simple definition, a mixture can be far more subtle than its mixed components. For instance, mixing gaussian laws may produce a potential with multiple deep wells. We study in the present work fine properties of mixtures with respect to concentration of measure and Sobolev type functional inequalities. We provide sharp Laplace bounds for Lipschitz functions in the case of generic mixtures, involving a transportation cost diameter of the mixed family. Additionally, our analysis of Sobolev type inequalities for two-component mixtures reveals natural relations with some kind of band isoperimetry and support constrained interpolation via mass transportation. We show that the Poincaré constant of a two-component mixture may remain bounded as the mixture proportion goes to 0 or 1 while the logarithmic Sobolev constant may surprisingly blow up. This counter-intuitive result is not reducible to support disconnections, and appears as a reminiscence of the variance-entropy comparison on the two-point space. As far as mixtures are concerned, the logarithmic Sobolev inequality is less stable than the Poincaré inequality and the sub-gaussian concentration for Lipschitz functions. We illustrate our results on a gallery of concrete two-component mixtures. This work leads to many open questions.

Les mélanges dont il est question ici sont des combinaisons convexes de lois de probabilité. Malgré cette définition simple, un mélange peut être beaucoup plus subtil que ses composants. Un mélange de lois gaussiennes par exemple peut donner lieu à des potentiels à profonds puits multiples. Dans ce travail, nous étudions les propriétés fines des mélanges vis à vis de la concentration de la mesure et des inégalités de type Sobolev. Nous proposons des bornes sur la transformée de Laplace faisant intervenir le diamètre de la famille mélangée pour une distance de transport. Notre analyse des inégalités de type Sobolev pour les mélanges à deux composants révèle des relations naturelles avec une forme d'isopérimétrie pour les bandes, ainsi qu'avec le transport optimal sous contrainte de support. Nous établissons que la constante de Poincaré peut rester bornée lorsque la proportion du mélange tend vers 0 tandis que la constante de Sobolev logarithmique peut exploser. Ce phénomène contre intuitif n'est pas réductible à un problème de support et peut être vu comme une trace de la comparaison variance-entropie sur l'espace à deux points. Pour les mélanges, la propriété de concentration de la mesure sous-gaussienne et l'inégalité de Poincaré sont plus stables que l'inégalité de Sobolev logarithmique. Nous illustrons nos résultats avec une collection d'exemples à deux composants concrets. Ce travail conduit à plusieurs questions ouvertes.

DOI : https://doi.org/10.1214/08-AIHP309
Classification:  60E15,  49Q20,  46E35,  62E99
Keywords: transportation cost distances, Mallows or Wasserstein distance, mixtures of distributions, finite gaussian mixtures, concentration of measure, gaussian bounds, tails probabilities, deviation inequalities, functional inequalities, Poincaré inequalities, Gross logarithmic Sobolev inequalities, band isoperimetry, transportation of measure, mass transportation
@article{AIHPB_2010__46_1_72_0,
     author = {Chafa\"\i , Djalil and Malrieu, Florent},
     title = {On fine properties of mixtures with respect to concentration of measure and Sobolev type inequalities},
     journal = {Annales de l'I.H.P. Probabilit\'es et statistiques},
     publisher = {Gauthier-Villars},
     volume = {46},
     number = {1},
     year = {2010},
     pages = {72-96},
     doi = {10.1214/08-AIHP309},
     zbl = {1204.60025},
     mrnumber = {2641771},
     language = {en},
     url = {http://www.numdam.org/item/AIHPB_2010__46_1_72_0}
}
Chafaï, Djalil; Malrieu, Florent. On fine properties of mixtures with respect to concentration of measure and Sobolev type inequalities. Annales de l'I.H.P. Probabilités et statistiques, Volume 46 (2010) no. 1, pp. 72-96. doi : 10.1214/08-AIHP309. http://www.numdam.org/item/AIHPB_2010__46_1_72_0/

[1] C. Ané, S. Blachère, D. Chafaï, P. Fougères, I. Gentil, F. Malrieu, C. Roberto and G. Scheffer. Sur les inégalités de Sobolev logarithmiques. Société Mathématique de France, Paris, 2000. (Preface by D. Bakry and M. Ledoux.) | MR 1845806 | Zbl 0982.46026

[2] D. Bakry and M. Émery. Diffusions hypercontractives. In Séminaire de probabilités (XIX, 1983/84) 177-206. Lecture Notes in Math. 1123. Springer, Berlin, 1985. | Numdam | MR 889476 | Zbl 0561.60080

[3] D. Bakry, M. Ledoux and F.-Y. Wang. Perturbations of functional inequalities using growth conditions. J. Math. Pures Appl. (9) 87 (2007) 394-407. | MR 2317340 | Zbl 1120.60070

[4] F. Barthe, P. Cattiaux and C. Roberto. Interpolated inequalities between exponential and Gaussian, Orlicz hypercontractivity and isoperimetry. Rev. Mat. Iberoamericana 22 (2006) 993-1067. | MR 2320410 | Zbl 1118.26014

[5] F. Barthe and C. Roberto. Sobolev inequalities for probability measures on the real line. Studia Math. 159 (2003) 481-497. (Dedicated to Professor Aleksander Pełczyński on the occasion of his 70th birthday (Polish).) | MR 2052235 | Zbl 1072.60008

[6] S. G. Bobkov. Concentration of normalized sums and a central limit theorem for noncorrelated random variables. Ann. Probab. 32 (2004) 2884-2907. | MR 2094433 | Zbl 1065.60006

[7] S. G. Bobkov. Generalized symmetric polynomials and an approximate de Finetti representation. J. Theoret. Probab. 18 (2005) 399-412. | MR 2137450 | Zbl 1079.28001

[8] S. G. Bobkov and F. Götze. Exponential integrability and transportation cost related to logarithmic Sobolev inequalities. J. Funct. Anal. 163 (1999) 1-28. | MR 1682772 | Zbl 0924.46027

[9] F. Bolley and C. Villani. Weighted Csiszár-Kullback-Pinsker inequalities and applications to transportation inequalities. Ann. Fac. Sci. Toulouse Math. 14 (2005) 331-352. | Numdam | MR 2172583 | Zbl 1087.60008

[10] L. A. Caffarelli. Monotonicity properties of optimal transportation and the FKG and related inequalities. Comm. Math. Phys. 214 (2000) 547-563. | MR 1800860 | Zbl 0978.60107

[11] L. A. Caffarelli. Erratum: [Comm. Math. Phys. 214 (2000) 547-563]. Comm. Math. Phys. 225 (2002) 449-450. | Zbl 0978.60107

[12] E. A. Carlen. Superadditivity of Fisher's information and logarithmic Sobolev inequalities. J. Funct. Anal. 101 (1991) 194-211. | MR 1132315 | Zbl 0732.60020

[13] S. Chatterjee. Spin glasses and Stein's method. Preprint, 2007. Available at arXiv:0706.3500v2 [math.PR]. | Zbl 1209.82016

[14] P. Diaconis and L. Saloff-Coste. Logarithmic Sobolev inequalities for finite Markov chains. Ann. Appl. Probab. 6 (1996) 695-750. | MR 1410112 | Zbl 0867.60043

[15] H. Djellout, A. Guillin and L. Wu. Transportation cost-information inequalities and applications to random dynamical systems and diffusions. Ann. Probab. 32 (2004) 2702-2732. | MR 2078555 | Zbl 1061.60011

[16] B. S. Everitt and D. J. Hand. Finite Mixture Distributions. Chapman & Hall, London, 1981. | MR 624267 | Zbl 0466.62018

[17] V. P. Fonf, J. Lindenstrauss and R. R. Phelps. Infinite dimensional convexity. In Handbook of the Geometry of Banach Spaces 599-670. North-Holland, Amsterdam, 2001. | MR 1863703 | Zbl 1086.46004

[18] S. Frühwirth-Schnatter. Finite Mixture and Markov Switching Models. Springer, New York, 2006. | MR 2265601 | Zbl 1108.62002

[19] I. Gentil. Inégalités de Sobolev logarithmique et de Poincaré pour la loi uniforme. Unpublished 2004. Available on the author's web page.

[20] I. Gentil and C. Roberto. Spectral gaps for spin systems: Some non-convex phase examples. J. Funct. Anal. 180 (2001) 66-84. | MR 1814423 | Zbl 0992.60091

[21] N. Gozlan. A characterization of dimension free concentration in terms of transportation inequalities. Preprint, 2008. Available at arXiv: 0804.3089 [math.PR]. | MR 2573565 | Zbl 1201.60016

[22] M. Gromov and V. D. Milman. A topological application of the isoperimetric inequality. Amer. J. Math. 105 (1983) 843-854. | MR 708367 | Zbl 0522.53039

[23] L. Gross. Logarithmic Sobolev inequalities. Amer. J. Math. 97 (1975) 1061-1083. | MR 420249 | Zbl 0318.46049

[24] L. Gross. Hypercontractivity, logarithmic Sobolev inequalities, and applications: A survey of surveys. In Diffusion, Quantum Theory, and Radically Elementary Mathematics 45-73. Math. Notes 47. Princeton Univ. Press, Princeton, NJ, 2006. | MR 2325763

[25] B. Helffer. Semiclassical Analysis, Witten Laplacians, and Statistical Mechanics. Series in Partial Differential Equations and Applications 1. World Scientific Publishing, River Edge, NJ, 2002. | MR 1936110 | Zbl 1046.82001

[26] W. Hoeffding. Probability inequalities for sums of bounded random variables. J. Amer. Statist. Assoc. 58 (1963) 13-30. | MR 144363 | Zbl 0127.10602

[27] R. Holley and D. Stroock. Logarithmic Sobolev inequalities and stochastic Ising models. J. Statist. Phys. 46 (1987) 1159-1194. | MR 893137 | Zbl 0682.60109

[28] M. Jerrum, J.-B. Son, P. Tetali and E. Vigoda. Elementary bounds on Poincaré and log-Sobolev constants for decomposable Markov chains. Ann. Appl. Probab. 14 (2004) 1741-1765. | MR 2099650 | Zbl 1067.60065

[29] O. Johnson. Convergence of the Poincaré constant. Teor. Veroyatnost. i Primenen. 48 (2003) 615-620. | MR 2141356 | Zbl 1054.60026

[30] I. Kontoyiannis and M. Madiman. Measure concentration for compound Poisson distributions. Electron. Comm. Probab. 11 (2006) 45-57. | MR 2219345 | Zbl 1112.60008

[31] L. Kontoyiannis and M. Madiman. Entropy, compound Poisson approximation, log-Sobolev inequalities and measure concentration. In Information Theory Workshop, 24-29 Oct. 2004. IEEE 71-75, 2004.

[32] R. Latała. On some inequalities for Gaussian measures. In Proceedings of the International Congress of Mathematicians, Vol. II Beijing, 2002 813-822. Higher Ed. Press, Beijing, 2002. | MR 1957087 | Zbl 1015.60011

[33] M. Ledoux. Concentration of measure and logarithmic Sobolev inequalities. In Séminaire de Probabilités, XXXIII 120-216. Springer, Berlin, 1999. | Numdam | MR 1767995 | Zbl 0957.60016

[34] M. Ledoux. The Concentration of Measure Phenomenon. Mathematical Surveys and Monographs 89. Amer. Math. Soc., Providence, RI, 2001. | MR 1849347 | Zbl 0995.60002

[35] N. Madras and D. Randall. Markov chain decomposition for convergence rate analysis. Ann. Appl. Probab. 12 (2002) 581-606. | MR 1910641 | Zbl 1017.60080

[36] K. Marton. A simple proof of the blowing-up lemma. IEEE Trans. Inform. Theory 32 (1986) 445-446. | MR 838213 | Zbl 0594.94003

[37] K. Marton. Bounding ̅d-distance by informational divergence: A method to prove measure concentration. Ann. Probab. 24 (1996) 857-866. | MR 1404531 | Zbl 0865.60017

[38] V. G. Maz'Ja. Sobolev Spaces. Springer, Berlin, 1985. (Translated from the Russian by T. O. Shaposhnikova.) | MR 817985 | Zbl 0692.46023

[39] G. Mclachlan and K. Basford. Mixture Models. Statistics: Textbooks and Monographs 84. Marcel Dekker Inc., New York, 1988. | MR 926484 | Zbl 0697.62050

[40] G. Mclachlan and D. Peel. Finite Mixture Models. Wiley, New York, 2000. | MR 1789474 | Zbl 0963.62061

[41] L Miclo. Quand est-ce que des bornes de Hardy permettent de calculer une constante de Poincaré exacte sur la droite?. Preprint, 2005. Available at http://hal.archives-ouvertes.fr/hal-00017875/en/. | Numdam | MR 2464097

[42] F. Otto and M. G. Reznikoff. A new criterion for the logarithmic Sobolev inequality and two applications. J. Funct. Anal. 243 (2007) 121-157. | MR 2291434 | Zbl 1109.60013

[43] R. R. Phelps, Lectures on Choquet's Theorem, 2nd edition. Lecture Notes in Math. 1757. Springer, Berlin, 2001. | MR 1835574 | Zbl 0997.46005

[44] S. T. Rachev. Probability Metrics and the Stability of Stochastic Models. Wiley, Chichester, 1991. | MR 1105086 | Zbl 0744.60004

[45] L. Saloff-Coste. Lectures on finite Markov chains. In Lectures on Probability Theory and Statistics (Saint-Flour, 1996) 301-413. Lecture Notes in Math. 1665. Springer, Berlin, 1997. | MR 1490046 | Zbl 0885.60061

[46] A. J. Stam. Some inequalities satisfied by the quantities of information of Fisher and Shannon. Information and Control 2 (1959) 101-112. | MR 109101 | Zbl 0085.34701

[47] V. N. Sudakov. Geometric problems in the theory of infinite-dimensional probability distributions. Proc. Steklov Inst. Math. (1979) 1-178. (Cover to cover translation of Trudy Mat. Inst. Steklov 141 (1976).) | MR 530375 | Zbl 0409.60005

[48] A. Takatsu. On Wasserstein geometry of the space of Gaussian measures, 2008. Available at arXiv:0801.2250 [math.DG].

[49] M. Talagrand. Transportation cost for Gaussian and other product measures. Geom. Funct. Anal. 6 (1996) 587-600. | MR 1392331 | Zbl 0859.46030

[50] D. M. Titterington, A. F. M. Smith and U. E. Makov. Statistical Analysis of Finite Mixture Distributions. Wiley, Chichester, 1985. | MR 838090 | Zbl 0646.62013

[51] C. Villani. Topics in Optimal Transportation. Graduate Studies in Mathematics 58. Amer. Math. Soc., Providence, RI, 2003. | MR 1964483 | Zbl 1106.90001

[52] C. Villani. Optimal Transport, Old and New. Springer, Berlin, 2009. | MR 2459454 | Zbl 1156.53003