Efficiency of pseudolikelihood estimation for simple Gaussian fields

By Julian Besag. 1977

How researchers see this paper

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

Citations in the corpus

Learning generative models for protein fold families (2011, methodology): “Instead of approximating the true-likelihood using approximate inference techniques, we use a different approximation based on a pseudo-likelihood proposed by [16], and used in [17, 18].”
Herding Dynamic Weights for Partially Observed Random Field Models (2009, methodology): “We show that under certain conditions averages computed over trajectories of the proposed dynamical system converge to averages computed over the data.”
Filters, Random Fields and Maximum Entropy (FRAME): Towards a Unified Theory for Texture Modeling (1998, background): “/ is as big as image I . often difﬁcult to discriminate when they produce similar marginal distributions for responses from a bank of ﬁlters ( Bergen and Adelson, 1991; Chubb and Landy, 1991).”
Flow-GAN: Combining Maximum Likelihood and Adversarial Learning in Generative Models (2017, methodology): “Hence, they are usually learned by approximating the likelihood using methods such as contrastive divergence (Carreira-Perpinan and Hinton 2005) and pseudolikelihood (Besag 1977).”
Estimation of Non-Normalized Statistical Models by Score Matching (2005, methodology): “Some consistency proofs were provided by Besag (1974, 1977), but these only apply to special cases such as Gaussian or binary random fields.”
Consistent estimation of the basic neighborhood of Markov random fields (2004, methodology): “As no maximum likelihood formula is available for MRF’s, here we use maximum pseudo-likelihood [ 1 ]; in the penalty term, a factor analogous but not equal to the number of free parameters will be used.”
Top-Down Learning for Structured Labeling with Convolutional Pseudoprior (2015, methodology): “Inspired by pseudo-likelihood [23] and the hybrid model in [31], we make an approximation to the posterior in Eq.”
Fast, Accurate, and Simple Models for Tabular Data via Augmented Distillation (2020, background): “For many models, maximum pseudolikelihood estimation produces asymptotically consistent parameter estimates, and often is more computationally tractable than optimizing the likelihood [23].”
Protein sequence design with a learned potential (2020, methodology): “The runtime for our method for sequence design is determined primarily by two steps: (1) sampling residues and rotamer angles and (2) computing the model energy (negative PLL).”
spatstat: An R Package for Analyzing Spatial Point Patterns (2005, background): “Disadvantages of maximum pseudolikelihood (MPL) include its small-sample bias and in-eﬃciency Besag (1977); Jensen and Møller (1991); Jensen and K¨unsch (1994) relative to maximum likelihood estimators (MLE).”
9. Neighborhood-Based Models for Social Networks (2002, methodology): “15 We rely on heuristic measures of fit since the distribution of –2LPL is not known (as for other applications of maximum pseudolikelihood estimation; see Besag, 1977, and Strauss & Ikeda, 1990).”
Decision tree fields (2011, background): “The pseudolikelihood [3] defines a surrogate likelihood function that is maximized.”
Deeply Learning the Messages in Message Passing Inference (2015, methodology): “Applying an approximate training objective is a solution to avoid repeat inference; pseudo-likelihood learning [6] and piecewise learning [7, 3] are examples of this kind of approach.”