The Annals of Applied Probability

The sample size required in importance sampling

Sourav Chatterjee and Persi Diaconis

Full-text: Access denied (no subscription detected)

We're sorry, but we are unable to provide you with the full text of this article because we are not able to identify you as a subscriber. If you have a personal subscription to this journal, then please login. If you are already logged in, then you may need to update your profile to register your subscription. Read more about accessing full-text

Abstract

The goal of importance sampling is to estimate the expected value of a given function with respect to a probability measure $\nu$ using a random sample of size $n$ drawn from a different probability measure $\mu$. If the two measures $\mu$ and $\nu$ are nearly singular with respect to each other, which is often the case in practice, the sample size required for accurate estimation is large. In this article, it is shown that in a fairly general setting, a sample of size approximately $\exp(D(\nu\parallel\mu))$ is necessary and sufficient for accurate estimation by importance sampling, where $D(\nu\parallel\mu)$ is the Kullback–Leibler divergence of $\mu$ from $\nu$. In particular, the required sample size exhibits a kind of cut-off in the logarithmic scale. The theory is applied to obtain a general formula for the sample size required in importance sampling for one-parameter exponential families (Gibbs measures).

Article information

Source
Ann. Appl. Probab., Volume 28, Number 2 (2018), 1099-1135.

Dates
Received: November 2015
Revised: June 2017
First available in Project Euclid: 11 April 2018

Permanent link to this document
https://projecteuclid.org/euclid.aoap/1523433632

Digital Object Identifier
doi:10.1214/17-AAP1326

Mathematical Reviews number (MathSciNet)
MR3784496

Zentralblatt MATH identifier
06897951

Subjects
Primary: 65C05: Monte Carlo methods 65C60: Computational problems in statistics 60F05: Central limit and other weak theorems 82B80: Numerical methods (Monte Carlo, series resummation, etc.) [See also 65-XX, 81T80]

Keywords
Importance sampling Monte Carlo methods Gibbs measure phase transition

Citation

Chatterjee, Sourav; Diaconis, Persi. The sample size required in importance sampling. Ann. Appl. Probab. 28 (2018), no. 2, 1099--1135. doi:10.1214/17-AAP1326. https://projecteuclid.org/euclid.aoap/1523433632


Export citation

References

  • [1] Agapiou, S., Papaspiliopoulos, O., Sanz-Alonso, D. and Stuart, A. M. (2017). Importance sampling: Computational complexity and intrinsic dimension. Preprint. Available at arXiv:1511.06196.
    arXiv: 1511.06196
  • [2] Asmussen, S. and Glynn, P. W. (2007). Stochastic Simulation: Algorithms and Analysis. Stochastic Modelling and Applied Probability 57. Springer, New York.
    Zentralblatt MATH: 1126.65001
  • [3] Bahadur, R. R. (1960). Some approximations to the binomial distribution function. Ann. Math. Stat. 31 43–54.
    Zentralblatt MATH: 0092.35203
    Digital Object Identifier: doi:10.1214/aoms/1177705986
    Project Euclid: euclid.aoms/1177705986
  • [4] Bassetti, F. and Diaconis, P. (2006). Examples comparing importance sampling and the Metropolis algorithm. Illinois J. Math. 50 67–91.
    Zentralblatt MATH: 1102.60060
    Project Euclid: euclid.ijm/1258059470
  • [5] Baxter, R. J. (1982). Exactly Solved Models in Statistical Mechanics. Academic Press, London.
    Zentralblatt MATH: 0538.60093
  • [6] Bhattacharya, B. B., Ganguly, S., Lubetzky, E. and Zhao, Y. (2015). Upper tails and independence polynomials in random graphs. Preprint. Available at arXiv:1507.04074.
    arXiv: 1507.04074
    Zentralblatt MATH: 1370.05099
    Digital Object Identifier: doi:10.1016/j.aim.2017.08.003
  • [7] Blanchet, J. and Glynn, P. (2008). Efficient rare-event simulation for the maximum of heavy-tailed random walks. Ann. Appl. Probab. 18 1351–1378.
    Zentralblatt MATH: 1147.60315
    Digital Object Identifier: doi:10.1214/07-AAP485
    Project Euclid: euclid.aoap/1216677125
  • [8] Blanchet, J., Glynn, P. and Leder, K. (2012). On Lyapunov inequalities and subsolutions for efficient importance sampling. ACM Trans. Model. Comput. Simul. 22 1104–1128.
  • [9] Blanchet, J. and Liu, J. (2008). State-dependent importance sampling for regularly varying random walks. Adv. Appl. Probab. 40 1104–1128.
    Zentralblatt MATH: 1159.60022
    Digital Object Identifier: doi:10.1239/aap/1231340166
    Project Euclid: euclid.aap/1231340166
  • [10] Blanchet, J. and Liu, J. (2010). Efficient importance sampling in ruin problems for multidimensional regularly varying random walks. J. Appl. Probab. 47 301–322.
    Zentralblatt MATH: 1257.60001
    Digital Object Identifier: doi:10.1239/jap/1276784893
    Project Euclid: euclid.jap/1276784893
  • [11] Blitzstein, J. and Diaconis, P. (2010). A sequential importance sampling algorithm for generating random graphs with prescribed degrees. Internet Math. 6 489–522.
    Zentralblatt MATH: 1238.60084
    Digital Object Identifier: doi:10.1080/15427951.2010.557277
    Project Euclid: euclid.im/1318514519
  • [12] Bousquet-Mélou, M. (2014). On the importance sampling of self-avoiding walks. Combin. Probab. Comput. 23 725–748.
    Zentralblatt MATH: 1298.05020
    Digital Object Identifier: doi:10.1017/S0963548313000643
  • [13] Cappé, O., Moulines, E. and Rydén, T. (2005). Inference in Hidden Markov Models. Springer, New York.
  • [14] Chan, H. P. and Lai, T. L. (2007). Efficient importance sampling for Monte Carlo evaluation of exceedance probabilities. Ann. Appl. Probab. 17 440–473.
    Zentralblatt MATH: 1134.65005
    Digital Object Identifier: doi:10.1214/105051606000000664
    Project Euclid: euclid.aoap/1174323253
  • [15] Chan, H. P. and Lai, T. L. (2011). A sequential Monte Carlo approach to computing tail probabilities in stochastic models. Ann. Appl. Probab. 21 2315–2342.
    Zentralblatt MATH: 1246.60042
    Digital Object Identifier: doi:10.1214/10-AAP758
    Project Euclid: euclid.aoap/1322057323
  • [16] Chatterjee, S. and Diaconis, P. (2013). Estimating and understanding exponential random graph models. Ann. Statist. 41 2428–2461.
    Zentralblatt MATH: 1293.62046
    Digital Object Identifier: doi:10.1214/13-AOS1155
    Project Euclid: euclid.aos/1383661269
  • [17] Chen, Y., Diaconis, P., Holmes, S. P. and Liu, J. S. (2005). Sequential Monte Carlo methods for statistical analysis of tables. J. Amer. Statist. Assoc. 100 109–120.
    Zentralblatt MATH: 1117.62310
    Digital Object Identifier: doi:10.1198/016214504000001303
  • [18] Chen, Y. and Liu, J. S. (2007). Sequential Monte Carlo methods for permutation tests on truncated data. Statist. Sinica 17 857–872.
    Zentralblatt MATH: 1133.62061
  • [19] Del Moral, P. (2004). Feynman–Kac Formulae: Genealogical and Interacting Particle Systems with Applications. Springer, New York.
    Zentralblatt MATH: 1130.60003
  • [20] Del Moral, P. (2013). Mean Field Simulation for Monte Carlo Integration. CRC Press, Boca Raton, FL.
    Zentralblatt MATH: 1282.65011
  • [21] Del Moral, P., Kohn, R. and Patras, F. (2015). A duality formula for Feynman–Kac path particle models. C. R. Math. Acad. Sci. Paris 353 465–469.
    Zentralblatt MATH: 1323.60126
    Digital Object Identifier: doi:10.1016/j.crma.2015.02.008
  • [22] Diaconis, P. and Zabell, S. (1991). Closed form summation for classical distributions: Variations on a theme of de Moivre. Statist. Sci. 6 284–302.
    Zentralblatt MATH: 0955.60500
    Digital Object Identifier: doi:10.1214/ss/1177011699
    Project Euclid: euclid.ss/1177011699
  • [23] Doucet, A., de Freitas, N. and Gordon, N., eds. (2001). Sequential Monte Carlo Methods in Practice. Springer, New York.
    Zentralblatt MATH: 0967.00022
  • [24] Dupuis, P., Spiliopoulos, K. and Wang, H. (2012). Importance sampling for multiscale diffusions. Multiscale Model. Simul. 10 1–27.
    Zentralblatt MATH: 1250.60031
    Digital Object Identifier: doi:10.1137/110842545
  • [25] Dupuis, P. and Wang, H. (2004). Importance sampling, large deviations, and differential games. Stoch. Stoch. Rep. 76 481–508.
    Zentralblatt MATH: 1076.65003
    Digital Object Identifier: doi:10.1080/10451120410001733845
  • [26] Efron, B. (2012). Bayesian inference and the parametric bootstrap. Ann. Appl. Stat. 6 1971–1997.
    Zentralblatt MATH: 1257.62027
    Digital Object Identifier: doi:10.1214/12-AOAS571
    Project Euclid: euclid.aoas/1356629067
  • [27] Freer, C. E., Mansinghka, V. K. and Roy, D. M. (2010). When are probabilistic programs probably computationally tractable? Presented at the NIPS Workshop on Monte Carlo Methods for Modern Applications, 2010. Available at http://danroy.org/papers/FreerManRoy-NIPSMC-2010.pdf.
  • [28] Gelman, A. and Meng, X.-L. (1998). Simulating normalizing constants: From importance sampling to bridge sampling to path sampling. Statist. Sci. 13 163–185.
    Zentralblatt MATH: 0966.65004
    Digital Object Identifier: doi:10.1214/ss/1028905934
    Project Euclid: euclid.ss/1028905934
  • [29] Hammersley, J. M. and Handscomb, D. C. (1965). Monte Carlo Methods. Methuen & Co., Ltd., London.
    Zentralblatt MATH: 0121.35503
  • [30] Hesterberg, T. (1995). Weighted average importance sampling and defensive mixture distributions. Technometrics 37 185–194.
    Zentralblatt MATH: 0822.62002
    Digital Object Identifier: doi:10.1080/00401706.1995.10484303
  • [31] Huggins, J. H. and Roy, D. M. (2015). Convergence of sequential Monte Carlo-based sampling methods. Preprint. Available at arXiv:1503.00966.
    arXiv: 1503.00966
  • [32] Hult, H. and Nyquist, P. (2016). Large deviations for weighted empirical measures arising in importance sampling. Stochastic Process. Appl. 126 138–170.
    Zentralblatt MATH: 1329.60053
    Digital Object Identifier: doi:10.1016/j.spa.2015.08.002
  • [33] Kahn, H. and Marshall, A. W. (1953). Methods of reducing sample size in Monte Carlo computations. J. Oper. Res. Soc. Am. 1 263–278.
  • [34] Kenyon, R., Kral, D., Radin, C. and Winkler, P. (2015). A variational principle for permutations. Preprint. Available at arXiv:1506.02340.
    arXiv: 1506.02340
  • [35] Kenyon, R., Radin, C., Ren, K. and Sadun, L. (2014). Multipodal structure and phase transitions in large constrained graphs. Preprint. Available at arXiv:1405.0599.
    arXiv: 1405.0599
    Zentralblatt MATH: 1376.82029
    Digital Object Identifier: doi:10.1007/s10955-017-1804-0
  • [36] Kenyon, R. and Yin, M. (2014). On the asymptotics of constrained exponential random graphs. Preprint. Available at arXiv:1406.3662.
    arXiv: 1406.3662
  • [37] Knuth, D. E. (1976). Mathematics and computer science: Coping with finiteness. Science 194 1235–1242.
    Zentralblatt MATH: 1225.68001
    Digital Object Identifier: doi:10.1126/science.194.4271.1235
  • [38] Knuth, D. E. (1996). Selected Papers on Computer Science. CSLI Lecture Notes 59. CSLI Publications, Stanford, CA; Cambridge University Press, Cambridge.
  • [39] Lelièvre, T., Rousset, M. and Stoltz, G. (2010). Free Energy Computations: A Mathematical Perspective. World Scientific, Singapore.
  • [40] Liu, J. S. (2008). Monte Carlo Strategies in Scientific Computing. Springer, New York.
    Zentralblatt MATH: 1132.65003
  • [41] Liu, J. S. and Chen, R. (1995). Blind deconvolution via sequential imputations. J. Amer. Statist. Assoc. 90 567–576.
    Zentralblatt MATH: 0826.62062
    Digital Object Identifier: doi:10.1080/01621459.1995.10476549
  • [42] Madras, N. (1998). Umbrella sampling and simulated tempering. In Numerical Methods for Polymeric Systems (Minneapolis, MN, 1996). IMA Vol. Math. Appl. 102 19–32. Springer, New York.
    Zentralblatt MATH: 0924.65147
  • [43] Madras, N. and Piccioni, M. (1999). Importance sampling for families of distributions. Ann. Appl. Probab. 9 1202–1225.
    Zentralblatt MATH: 0966.60061
    Digital Object Identifier: doi:10.1214/aoap/1029962870
    Project Euclid: euclid.aoap/1029962870
  • [44] McCoy, B. M. (2010). Advanced Statistical Mechanics. International Series of Monographs on Physics 146. Oxford Univ. Press, Oxford.
  • [45] Mukherjee, S. (2013). Estimation in exponential families on permutations. Preprint. Available at arXiv:1307.0978.
    arXiv: 1307.0978
  • [46] Naiman, D. Q. and Wynn, H. P. (1997). Abstract tubes, improved inclusion-exclusion identities and inequalities and importance sampling. Ann. Statist. 25 1954–1983.
    Zentralblatt MATH: 0902.60017
    Digital Object Identifier: doi:10.1214/aos/1069362380
    Project Euclid: euclid.aos/1069362380
  • [47] Owen, A. and Zhou, Y. (1999). Adaptive importance sampling by mixtures of products of beta distributions. Technical report No. 1999–25, Dept. Statistics, Stanford Univ., Stanford, CA.
  • [48] Owen, A. and Zhou, Y. (2000). Safe and effective importance sampling. J. Amer. Statist. Assoc. 95 135–143.
    Zentralblatt MATH: 0998.65003
    Digital Object Identifier: doi:10.1080/01621459.2000.10473909
  • [49] Owen, A. B. (2005). Multidimensional variation for quasi-Monte Carlo. In Contemporary Multivariate Analysis and Design of Experiments. Ser. Biostat. 2 49–74. World Sci. Publ., Hackensack, NJ.
    Zentralblatt MATH: 1266.26024
  • [50] Owen, A. B. (2006). Quasi-Monte Carlo for integrands with point singularities at unknown locations. In Monte Carlo and Quasi-Monte Carlo Methods 2004 403–417. Springer, Berlin.
    Zentralblatt MATH: 1097.65006
  • [51] Robert, C. P. and Casella, G. (2004). Monte Carlo Statistical Methods, 2nd ed. Springer, New York.
    Zentralblatt MATH: 1096.62003
  • [52] Rosenbluth, M. N. and Rosenbluth, A. W. (1955). Monte Carlo calculation of the average extension of molecular chains. J. Chem. Phys. 23 356–359.
  • [53] Shi, J., Siegmund, D. and Yakir, B. (2007). Importance sampling for estimating $p$ values in linkage analysis. J. Amer. Statist. Assoc. 102 929–937.
    Zentralblatt MATH: 05564421
    Digital Object Identifier: doi:10.1198/016214507000000680
  • [54] Siegmund, D. (1976). Importance sampling in the Monte Carlo study of sequential tests. Ann. Statist. 4 673–684.
    Zentralblatt MATH: 0353.62044
    Digital Object Identifier: doi:10.1214/aos/1176343541
    Project Euclid: euclid.aos/1176343541
  • [55] Srinivasan, R. (2002). Importance Sampling: Applications in Communications and Detection. Springer, Berlin.
    Zentralblatt MATH: 1013.62003
  • [56] Starr, S. (2009). Thermodynamic limit for the Mallows model on $S_{n}$. J. Math. Phys. 50 095208.
  • [57] Torrie, G. M. and Valleau, J. P. (1977). Nonphysical sampling distributions in Monte Carlo free-energy estimation: Umbrella sampling. J. Comput. Phys. 23 187–199.
  • [58] Whiteley, N., Lee, A. and Heine, K. (2016). On the role of interaction in sequential Monte Carlo algorithms. Bernoulli 22 494–529.
    Zentralblatt MATH: 06543278
    Digital Object Identifier: doi:10.3150/14-BEJ666
    Project Euclid: euclid.bj/1443620858

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