The Annals of Applied Probability

Small-world MCMC and convergence to multi-modal distributions: From slow mixing to fast mixing

Yongtao Guan and Stephen M. Krone

Full-text: Open access

Enhanced PDF (198 KB)

Abstract

We compare convergence rates of Metropolis–Hastings chains to multi-modal target distributions when the proposal distributions can be of “local” and “small world” type. In particular, we show that by adding occasional long-range jumps to a given local proposal distribution, one can turn a chain that is “slowly mixing” (in the complexity of the problem) into a chain that is “rapidly mixing.” To do this, we obtain spectral gap estimates via a new state decomposition theorem and apply an isoperimetric inequality for log-concave probability measures. We discuss potential applicability of our result to Metropolis-coupled Markov chain Monte Carlo schemes.

Article information

Source
Ann. Appl. Probab., Volume 17, Number 1 (2007), 284-304.

Dates
First available in Project Euclid: 13 February 2007

Permanent link to this document
https://projecteuclid.org/euclid.aoap/1171377185

Digital Object Identifier
doi:10.1214/105051606000000772

Mathematical Reviews number (MathSciNet)
MR2292588

Zentralblatt MATH identifier
1139.65001

Subjects
Primary: 65C05: Monte Carlo methods
Secondary: 65C40: Computational Markov chains

Keywords
Markov chain Monte Carlo small world spectral gap Cheeger’s inequality state decomposition isoperimetric inequality Metropolis-coupled MCMC

Citation

Guan, Yongtao; Krone, Stephen M. Small-world MCMC and convergence to multi-modal distributions: From slow mixing to fast mixing. Ann. Appl. Probab. 17 (2007), no. 1, 284--304. doi:10.1214/105051606000000772. https://projecteuclid.org/euclid.aoap/1171377185


Export citation

References

  • Applegate, D. and Kannan, R. (1990). Sampling and integration of near log-concave functions. In Proc. 23rd ACM STOC 156--163. ACM Press, New York.
  • Bhatnagar, N. and Randall, D. (2004). Torpid mixing of simulated tempering on the Potts model. In Proceedings of the 15th Annual ACM--SIAM Symposium on Discrete Algorithms (New Orleans, LA) 478--487. SIAM, Philadelphia.
    Mathematical Reviews (MathSciNet): MR2291087
  • Bobkov, S. G. (1999). Isoperimetric and analytic inequalities for log-concave probability measures. Ann. Probab. 27 1903--1921.
    Mathematical Reviews (MathSciNet): MR1742893
    Digital Object Identifier: doi:10.1214/aop/1022677553
    Project Euclid: euclid.aop/1022874820
    Zentralblatt MATH: 0964.60013
  • Borell, C. (1974). Convex measures on locally convex spaces. Ark. Math. 12 239--252.
    Mathematical Reviews (MathSciNet): MR0388475
    Digital Object Identifier: doi:10.1007/BF02384761
    Zentralblatt MATH: 0297.60004
  • Chan, K. S. and Geyer, C. J. (1994). Discussion of the paper by Tierney. Ann. Statist. 22 1747--1758.
  • Geyer, C. J. (1991). Markov chain Monte Carlo maximum likelihood. In Computing Science and Statistics: Proceedings of the 23rd Symposium on the Interface (E. M. Keramides, ed.) 156--163. Interface Foundation, Fairfax Station.
    Zentralblatt MATH: 0751.12004
  • Geyer, C. J. and Thompson, E. A. (1995). Annealing Markov chain Monte Carlo with applications to ancestral inference. J. Amer. Statist. Assoc. 90 909--920.
  • Guan, Y., Flei$\ss$ner, R., Joyce, P. and Krone, S. M. (2006). Markov chain Monte Carlo in small worlds. Statist. Comput. 16 193--202.
    Mathematical Reviews (MathSciNet): MR2227395
    Digital Object Identifier: doi:10.1007/s11222-006-6966-6
  • Hastings, W. K. (1970). Monte Carlo sampling methods using Markov Chains and their applications. Biometrika 57 97--109.
  • Jarner, S. F. and Yuen, W. K. (2004). Conductance bounds on the $L^2$ convergence rate of Metropolis algorithms on unbounded state spaces. Adv. in Appl. Probab. 36 243--266.
    Mathematical Reviews (MathSciNet): MR2035782
    Digital Object Identifier: doi:10.1239/aap/1077134472
    Project Euclid: euclid.aap/1077134472
    Zentralblatt MATH: 1042.60040
  • Kannan, R. and Li, G. (1996). Sampling according to the multivariate normal density. In 37th Annual IEEE Symposium on Foundations of Computer Science (FOCS'96) 204--212. IEEE Comput. Soc. Press, Los Alamitos, CA.
    Mathematical Reviews (MathSciNet): MR1450618
  • Kannan, R., Lovász, L. and Simonovits, M. (1995). Isoperimetric problems for convex bodies and a localization lemma. Discrete Comput. Geom. 13 541--559.
    Mathematical Reviews (MathSciNet): MR1318794
  • Kirkpatrick, S., Jr., Gelatt, C. D. and Vecchi, M. P. (1983). Optimization by simulated annealing. Science 220 671--680.
    Mathematical Reviews (MathSciNet): MR0702485
    Digital Object Identifier: doi:10.1126/science.220.4598.671
  • Lawler, G. F. and Sokal, A. D. (1988). Bounds on the $L^2$ spectrum for Markov chains and Markov processes: A generalization of Cheeger's inequality. Trans. Amer. Math. Soc. 309 557--580.
    Mathematical Reviews (MathSciNet): MR0930082
    Digital Object Identifier: doi:10.2307/2000925
    JSTOR: links.jstor.org
    Zentralblatt MATH: 0716.60073
  • Leindler, L. (1972). On a certain converse of Hölder's inequality. II. Acta Sci. Math. (Szeged) 33 217--223.
    Mathematical Reviews (MathSciNet): MR2199372
  • Lovász, L. and Simonovits, M. (1993). Random walks in a convex body and an improved volume algorithm. Random Structures Algorithms 4 359--412.
    Mathematical Reviews (MathSciNet): MR1238906
  • Lovász, L. and Vempala, S. (2003). The geometry of logconcave functions and an $O^*(n^3)$ sampling algorithm. Microsoft Research Technical Report MSR-TR-2003--4. Available at http://www-math.mit.edu/~vempala/papers/logcon-ball.pdf.
  • Lovász, L. and Vempala, S. (2003). Logconcave functions: Geometry and efficient sampling algorithm. In 44th Annual IEEE Symposium on Foundations of Computer Science (FOCS'03) 1--10. IEEE Comput. Soc. Press, Los Alamitos, CA.
  • Madras, N. and Randall, D. (2002). Markov chain decomposition for convergence rate analysis. Ann. Appl. Probab. 12 581--606.
    Mathematical Reviews (MathSciNet): MR1910641
    Digital Object Identifier: doi:10.1214/aoap/1026915617
    Project Euclid: euclid.aoap/1026915617
    Zentralblatt MATH: 1017.60080
  • Marinari, E. and Parisi, G. (1992). Simulated tempering: A new Monte Carlo scheme. Europhys. Lett. 19 451--458.
  • Martin, R. A. and Randall, D. (2000). Sampling adsorbing staircase walks using a new Markov chain decomposition method. In 41st Annual IEEE Symposium on Foundations of Computer Science (FOCS'01) 492--502. IEEE Comput. Soc. Press, Los Alamitos, CA.
    Mathematical Reviews (MathSciNet): MR1931846
  • Metropolis, N., Rosenbluth, A. E., Rosenbluth, M. N., Teller, A. H. and Teller, E. (1953). Equation of state calculations by fast computing machines. J. Chem. Phys. 21 1087--1091.
  • Peña, J. M. (2005). Exclusion and inclusion intervals for the real eigenvalues of positive matrices. SIAM J. Matrix Anal. Appl. 26 908--917.
    Mathematical Reviews (MathSciNet): MR2178204
    Digital Object Identifier: doi:10.1137/04061074X
    Zentralblatt MATH: 1082.15031
  • Prékopa, A. (1973). Logarithmic concave measures and functions. Acta Sci. Math. (Szeged) 34 335--343.
    Mathematical Reviews (MathSciNet): MR0404557
  • Roberts, G. O. and Rosenthal, J. S. (1997). Geometric ergodicity and hybrid Markov chains. Electron. Comm. Probab. 2 13--25.
    Mathematical Reviews (MathSciNet): MR1448322
  • Roberts, G. O. and Tweedie, R. L. (2001). Geometric $L^2$ and $L^1$ convergence are equivalent for reversible Markov chains. J. Appl. Probab. 38 37--41.
    Mathematical Reviews (MathSciNet): MR1915532
    Digital Object Identifier: doi:10.1239/jap/1085496589
    Project Euclid: euclid.jap/1085496589
    Zentralblatt MATH: 1011.60050
  • Watts, D. J. and Strogatz, S. H. (1998). Collective dynamics of `small-world' networks. Nature 393 440--442.

The Institute of Mathematical Statistics

The Annals of Applied Probability

New content alerts

Email RSS ToC RSS Article
  • You have access to this content.
  • You have partial access to this content.
  • You do not have access to this content.
Turn Off MathJax

What is MathJax?

More like this

+ See more