Non-linear regression models for Approximate Bayesian Computation

被引:318
作者
Blum, Michael G. B. [1 ]
Francois, Olivier [2 ]
机构
[1] Fac Med Grenoble, TIMC IMAG, CNRS, F-38706 La Tronche, France
[2] Fac Med Grenoble, TIMC IMAG, Inst Natl Polytech Grenoble, F-38706 La Tronche, France
关键词
Likelihood-free inference; Curse of dimensionality; Feed forward neural networks; Heteroscedasticity; Coalescent models; Approximate Bayesian computation; Conditional density estimation; Implicit statistical models; Importance sampling; Non-linear regression; Indirect inference; HUMAN-POPULATION EXPANSION; MONTE-CARLO METHODS; MAXIMUM-LIKELIHOOD; INFERENCE; STATISTICS; SUPPORT;
D O I
10.1007/s11222-009-9116-0
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
Approximate Bayesian inference on the basis of summary statistics is well-suited to complex problems for which the likelihood is either mathematically or computationally intractable. However the methods that use rejection suffer from the curse of dimensionality when the number of summary statistics is increased. Here we propose a machine-learning approach to the estimation of the posterior density by introducing two innovations. The new method fits a nonlinear conditional heteroscedastic regression of the parameter on the summary statistics, and then adaptively improves estimation using importance sampling. The new algorithm is compared to the state-of-the-art approximate Bayesian methods, and achieves considerable reduction of the computational burden in two examples of inference in statistical genetics and in a queueing model.
引用
收藏
页码:63 / 73
页数:11
相关论文
共 51 条
[21]  
Hardle W., 2004, Nonparametric and semiparametric models
[22]   Estimating functions in indirect inference [J].
Heggland, K ;
Frigessi, A .
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-STATISTICAL METHODOLOGY, 2004, 66 :447-462
[23]   Integration within the Felsenstein equation for improved Markov chain Monte Carlo methods in population genetics [J].
Hey, Jody ;
Nielsen, Rasmus .
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA, 2007, 104 (08) :2785-2790
[24]   A power analysis of microsatellite-based statistics for inferring past population growth [J].
King, JP ;
Kimmel, M ;
Chakraborty, R .
MOLECULAR BIOLOGY AND EVOLUTION, 2000, 17 (12) :1859-1868
[25]   LAMARC 2.0: maximum likelihood and Bayesian estimation of population parameters [J].
Kuhner, MK .
BIOINFORMATICS, 2006, 22 (06) :768-770
[26]  
Liu J.S., 2001, Monte Carlo Strategies in Scientific Computing
[27]   Markov chain Monte Carlo without likelihoods [J].
Marjoram, P ;
Molitor, J ;
Plagnol, V ;
Tavaré, S .
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA, 2003, 100 (26) :15324-15328
[28]   Modern computational approaches for analysing molecular genetic variation data [J].
Marjoram, Paul ;
Tavare, Simon .
NATURE REVIEWS GENETICS, 2006, 7 (10) :759-770
[29]  
Nadaraya E.A., 1964, Theory Prob. Appl, V9, P141, DOI DOI 10.1137/1109020
[30]  
Nix D. A., 1995, Advances in Neural Information Processing Systems 7, P489