Convergence of Adaptive Sampling Schemes

In the design of ecient simulation algorithms, one is often beset with a poor choice of proposal distributions. Although the performances of a given kernel can clarify how adequate it is for the problem at hand, a permanent on-line modification of kernels causes concerns about the validity of the resulting algorithm. While the issue is quite complex and most often intractable for MCMC algorithms, the equivalent version for importance sampling algorithms can be validated quite precisely. We derive sucient convergence conditions for a wide class of population Monte Carlo algorithms and show that Rao‐ Blackwellized versions asymptotically achieve an optimum in terms of a Kullback divergence criterion, while more rudimentary versions simply do not benefit from repeated updating.

[1]  R. Tweedie,et al.  Rates of convergence of the Hastings and Metropolis algorithms , 1996 .

[2]  D. Rubin.  Using the SIR algorithm to simulate posterior distributions , 1988 .

[3]  Reuven Y. Rubinstein,et al.  Simulation and the Monte Carlo method , 1981, Wiley series in probability and mathematical statistics.

[4]  Jean-Michel Marin,et al.  Iterated importance sampling in missing data problems , 2006, Comput. Stat. Data Anal..

[5]  L. Tierney.  Markov Chains for Exploring Posterior Distributions , 1994 .

[6]  A. Gelman,et al.  Weak convergence and optimal scaling of random walk Metropolis algorithms , 1997 .

[7]  Heikki Haario,et al.  Adaptive proposal distribution for random walk Metropolis algorithm , 1999, Comput. Stat..

[8]  P. Moral,et al.  Sequential Monte Carlo samplers , 2002, cond-mat/0212648.

[9]  Eric P. Fox.  Bayesian Statistics 3 , 1991 .

[10]  Hoon Kim,et al.  Monte Carlo Statistical Methods , 2000, Technometrics.

[11]  C. Andrieu,et al.  On the ergodicity properties of some adaptive MCMC algorithms , 2006, math/0610317.

[12]  Donald B. Rubin,et al.  Comment : A noniterative sampling/importance resampling alternative to the data augmentation algorithm for creating a few imputations when fractions of missing information are modest : The SIR Algorithm , 1987 .

[13]  D. Rubin,et al.  The calculation of posterior distributions by data augmentation , 1987 .

[14]  Eric Moulines,et al.  Inference in hidden Markov models , 2010, Springer series in statistics.

[15]  Yukito Iba,et al.  Population-based Monte Carlo algorithms , 2000 .

[16]  H. Kunsch.  Recursive Monte Carlo filters: Algorithms and theoretical analysis , 2006, math/0602211.

[17]  H. Haario,et al.  An adaptive Metropolis algorithm , 2001 .

[18]  Brian D. Ripley,et al.  Stochastic Simulation , 2005 .

[19]  J. Marin,et al.  Population Monte Carlo , 2004 .

[20]  G. Roberts,et al.  Adaptive Markov Chain Monte Carlo through Regeneration , 1998 .

[21]  Anatoly Zhigljavsky,et al.  Self-regenerative Markov chain Monte Carlo with adaptation , 2003 .

[22]  Timothy J. Robinson,et al.  Sequential Monte Carlo Methods in Practice , 2003 .

[23]  N. Chopin.  Central limit theorem for sequential Monte Carlo methods and its application to Bayesian inference , 2004, math/0508594.

[24]  T. Hesterberg,et al.  Weighted Average Importance Sampling and Defensive Mixture Distributions , 1995 .

[25]  Haikady N. Nagaraja,et al.  Inference in Hidden Markov Models , 2006, Technometrics.

[26]  Brian Jefferies.  Feynman-Kac Formulae , 1996 .