Advertisement

Statistics and Computing

, Volume 20, Issue 1, pp 1–7 | Cite as

Importance tempering

Article

Abstract

Simulated tempering (ST) is an established Markov chain Monte Carlo (MCMC) method for sampling from a multimodal density π(θ). Typically, ST involves introducing an auxiliary variable k taking values in a finite subset of [0,1] and indexing a set of tempered distributions, say π k (θ) π(θ) k . In this case, small values of k encourage better mixing, but samples from π are only obtained when the joint chain for (θ,k) reaches k=1. However, the entire chain can be used to estimate expectations under π of functions of interest, provided that importance sampling (IS) weights are calculated. Unfortunately this method, which we call importance tempering (IT), can disappoint. This is partly because the most immediately obvious implementation is naïve and can lead to high variance estimators. We derive a new optimal method for combining multiple IS estimators and prove that the resulting estimator has a highly desirable property related to the notion of effective sample size. We briefly report on the success of the optimal combination in two modelling scenarios requiring reversible-jump MCMC, where the naïve approach fails.

Keywords

Simulated tempering Importance sampling Markov chain Monte Carlo (MCMC) Metropolis–coupled MCMC 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. Atchadé, Y., Liu, J.: The Wang–Landau algorithm in general state spaces: applications and convergence analysis. Technical Report, University of Harvard (2007) Google Scholar
  2. Breiman, L., Friedman, J.H., Olshen, R., Stone, C.: Classification and Regression Trees. Wadsworth, Belmont (1984) MATHGoogle Scholar
  3. Del Moral, P., Doucet, A., Jasra, A.: Sequential Monte Carlo samplers. J. R. Stat. Soc. Ser. B 68, 411–436 (2006) MATHCrossRefMathSciNetGoogle Scholar
  4. Geyer, C.: Markov chain Monte Carlo maximum likelihood. In: Computing Science and Statistics: Proceedings of the 23rd Symposium on the Interface, pp. 156–163 (1991) Google Scholar
  5. Geyer, C., Thompson, E.: Annealing Markov chain Monte Carlo with applications to ancenstral inference. J. Am. Stat. Assoc. 90, 909–920 (1995) MATHCrossRefGoogle Scholar
  6. Gramacy, R.B., Lee, H.K.H.: Bayesian treed Gaussian process models with an application to computer modeling. J. Am. Stat. Assoc. 103(143), 1119–1130 (2008) CrossRefGoogle Scholar
  7. Green, P.: Reversible jump Markov chain Monte Carlo computation and Bayesian model determination. Biometrika 82, 711–732 (1995) MATHCrossRefMathSciNetGoogle Scholar
  8. Hukushima, K., Nemoto, K.: Exchange Monte Carlo method and application to spin glass simulations. J. Phys. Soc. Jpn. 65(4), 1604–1608 (1996) CrossRefGoogle Scholar
  9. Iba, Y.: Extended ensemble Monte Carlo. Int. J. Mod. Phys. 12(5), 623–656 (2001) CrossRefGoogle Scholar
  10. Jasra, A., Stephens, D., Holmes, C.: On population-based simulation for static inference. Stat. Comput. 17(3), 263–279 (2007a) CrossRefMathSciNetGoogle Scholar
  11. Jasra, A., Stephens, D., Holmes, C.: Population-based reversible jump Markov chain Monte Carlo. Biometrika 94(4), 787–807 (2007b) MATHCrossRefMathSciNetGoogle Scholar
  12. Jennison, C.: Discussion on the meeting on the Gibbs sampler and other Markov chain Monte Carlo methods. J. R. Stat. Soc. Ser. B 55, 54–56 (1993) Google Scholar
  13. Kass, R.E., Carlin, B.P., Gelman, A., Neal, R.M.: Markov chain Monte Carlo in practice: a roundtable discussion. Am. Stat. 52(2), 93–100 (1998) CrossRefMathSciNetGoogle Scholar
  14. King, R., Brooks, S.: Model selection for integrated recovery/recapture data. Biometrics 58, 841–851 (2002) CrossRefMathSciNetGoogle Scholar
  15. Kirkpatrick, S., Gelatt, C., Vecci, M.: Optimization by simulated annealing. Science 220, 671–680 (1983) CrossRefMathSciNetGoogle Scholar
  16. Kushner, H., Lin, G.: Stochastic Approximation Algorithms and Applications. Springer, New York (1997) MATHGoogle Scholar
  17. Liu, J.S.: Monte Carlo Strategies in Scientific Computing. Springer, New York (2001) MATHGoogle Scholar
  18. Madras, N., Picconi, M.: Importance sampling for families of distributions. Ann. Appl. Probab. 9, 1202–1225 (1999) MATHCrossRefMathSciNetGoogle Scholar
  19. Marinari, E., Parisi, G.: Simulated tempering: a new Monte Carlo scheme. Europhys. Lett. 19, 451–458 (1992) CrossRefGoogle Scholar
  20. Neal, R.M.: Sampling from multimodal distributions using tempered transition. Stat. Comput. 6, 353–366 (1996) CrossRefGoogle Scholar
  21. Neal, R.M.: Annealed importance sampling. Stat. Comput. 11, 125–129 (2001) CrossRefMathSciNetGoogle Scholar
  22. Neal, R.M.: Estimating ratios of normalizing constants using Linked Importance Sampling. Technical Report 0511, Department of Statistics, University of Toronto (2005), 37 pages Google Scholar
  23. Owen, A., Zhou, Y.: Safe and effective importance sampling. J. Am. Stat. Assoc. 95(449), 135–143 (2000) MATHCrossRefMathSciNetGoogle Scholar
  24. Veach, E., Guibas, L.J.: Optimally combining sampling techniques for Monte Carlo rendering. In: SIGGRAPH ’95 Conference Proceedings, pp. 419–428. Reading, Addison-Wesley (1995) Google Scholar
  25. Wong, W., Liang, F.: (1997). Dynamic weighting in Monte Carlo and optimization. In: Proceedings of the National Academy of Sciences of USA, vol. 94(26), pp. 14220–14224 Google Scholar

Copyright information

© Springer Science+Business Media, LLC 2008

Authors and Affiliations

  1. 1.Statistical LaboratoryUniversity of CambridgeCambridgeUK
  2. 2.CREEMUniversity of St. AndrewsSt. AndrewsScotland

Personalised recommendations