Conditionally structured variational Gaussian approximation with importance weights
From MaRDI portal
Publication:2209703
Abstract: We develop flexible methods of deriving variational inference for models with complex latent variable structure. By splitting the variables in these models into "global" parameters and "local" latent variables, we define a class of variational approximations that exploit this partitioning and go beyond Gaussian variational approximation. This approximation is motivated by the fact that in many hierarchical models, there are global variance parameters which determine the scale of local latent variables in their posterior conditional on the global parameters. We also consider parsimonious parametrizations by using conditional independence structure, and improved estimation of the log marginal likelihood and variational density using importance weights. These methods are shown to improve significantly on Gaussian variational approximation methods for a similar computational cost. Application of the methodology is illustrated using generalized linear mixed models and state space models.
Recommendations
- Variational approximation for importance sampling
- Gaussian Variational Approximation With a Factor Covariance Structure
- Importance sampling as a variational approximation
- Variational Bayesian inference with Gaussian-mixture approximations
- Variational inference in nonconjugate models
- Variational inference and sparsity in high-dimensional deep Gaussian mixture models
- Gauging variational inference
- The Variational Gaussian Approximation Revisited
Cites work
- \(\alpha\)-variational inference with statistical guarantees
- A general framework for the parametrization of hierarchical models
- A likelihood-based method for analysing longitudinal binary responses
- A new approach to Cholesky-based covariance regularization in high dimensions
- Ancillarity-sufficiency interweaving strategy (ASIS) for boosting MCMC estimation of stochastic volatility models
- Approximate Inference in Generalized Linear Mixed Models
- Explaining variational approximations
- Fixed-form variational posterior approximation through stochastic linear regression
- Gaussian variational approximation with sparse precision matrices
- scientific article; zbMATH DE number 3173999 (Why is no real title available?)
- scientific article; zbMATH DE number 1222284 (Why is no real title available?)
- scientific article; zbMATH DE number 1261669 (Why is no real title available?)
- Inference via low-dimensional couplings
- Introduction to Stochastic Search and Optimization
- Rényi Divergence and Kullback-Leibler Divergence
- Some Covariance Models for Longitudinal Count Data with Overdispersion
- The Elimination Matrix: Some Lemmas and Applications
- Variational inference for generalized linear mixed models using partially noncentered parametrizations
Cited in
(8)- Gaussian variational approximation with sparse precision matrices
- A variational inference for the Lévy adaptive regression with multiple kernels
- Variational inference for generalized linear mixed models using partially noncentered parametrizations
- Natural gradient hybrid variational inference with application to deep mixed models
- Gaussian variational approximations for high-dimensional state space models
- Flexible Variational Bayes Based on a Copula of a Mixture
- Structured Variational Approximations with Skew Normal Decomposable Graphical Models and Implicit Copulas
- Variational Inference for Hierarchical Models with Conditional Scale and Skewness Corrections
This page was built for publication: Conditionally structured variational Gaussian approximation with importance weights
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2209703)
