To Center or Not to Center: That Is Not the Question—An Ancillarity–Sufficiency Interweaving Strategy (ASIS) for Boosting MCMC Efficiency

By Yaming Yu, Xiao-Li Meng. 2011

242 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Statistics and Probability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Statistics and Probability

Citations in the corpus

Unbiased Markov chain Monte Carlo methods with couplings (2020, methodology): “…j = i} Step 3: compute the final estimator H̄Åk:m as the (sample) average of {HÅk:m.X.i/, Y.i/;ηi/}Ni=1 via efficiency swindles (e.g. Van Dyk and Meng (2001), Craiu and Meng (2001, 2005), Craiu and Lemieux (2007) and Yu and Meng (2011)), among which figures prominently the control variates method.”
Analysis of the Gibbs Sampler for Hierarchical Inverse Problems (2013, methodology): “We have considered the interweaving method of [41], which combines in each iteration centered and non-centered draws of δ , and the partially non-centered parametrizations of [33], in which the prior is reparametrized as u = δ − t 2 v t where v t ∼ N (0 , δ t − 1 C 0 ), for some t ∈ [0 , 1].”
Pseudo-Marginal Bayesian Inference for Gaussian Processes (2013, methodology): “In this parameterization, that is also known as Ancillary Augmentation (AA) [36], ν are constructed to be a priori independent from the hyper-parameters (using L is convenient as it is needed also to evaluate the GP prior density).”
Sparse Bayesian time-varying covariance estimation in many dimensions (2016, methodology): “3 (2)We use the C-level interface to update in the R (R Core Team 2016) package stochvol (Kastner 2016) which provides a single sweep of the draws implementing an efficient AWOL (Kastner and FrühwirthSchnatter 2014) sampler with interweaving (Yu and Meng 2011).”
Compound random measures and their use in Bayesian non‐parametrics (2014, methodology): “, vd uses a variation on the interweaving approach of [38], which leads to better mixing than the standard full conditional distribution for vj .”
Variational Inference for Generalized Linear Mixed Models Using Partially Noncentered Parametrizations (2012, methodology): “In this section we present the nonconjugate variational message passing algorithm recently developed in machine learning by Knowles and Minka (2011) for ﬁtting GLMMs. Recall that in VB, the posterior distribution p ( θ | y ) is approximated by a q ( θ ) which is assumed to be of a factorized form, s”