Abstract:Birgir Hrafnkelsson, Stefan Siegert, Raphaël Huser, Haakon Bakka, Árni V. Jóhannesson. With modern high-dimensional data, complex statistical models are necessary, requiring computationally feasible inference schemes. We introduce Max-and-Smooth, an approximate Bayesian inference scheme for a flexible class of latent Gaussian models (LGMs) where one or more of the likelihood parameters are modeled by latent additive Gaussian processes. Our proposed inference scheme is a two-step approach. In the first step (Max), the likelihood function is approximated by a Gaussian density with mean and covariance equal to either (a) the maximum likelihood estimate and the inverse observed information, respectively, or (b) the mean and covariance of the normalized likelihood function. In the second step (Smooth), the latent parameters and hyperparameters are inferred and smoothed with the approximated likelihood function. The proposed method ensures that the uncertainty from the first step is correctly propagated to the second step. Because the prior density for the latent parameters is assumed to be Gaussian and the approximated likelihood function is Gaussian, the approximate posterior density of the latent parameters (conditional on the hyperparameters) is also Gaussian, thus facilitating efficient posterior inference in high dimensions. Furthermore, the approximate marginal posterior distribution of the hyperparameters is tractable, and as a result, the hyperparameters can be sampled independently of the latent parameters. We show that the computational cost of Max-and-Smooth is close to being insensitive to the number of independent data replicates, and that it scales well with increased dimension of the latent parameter vector provided that its Gaussian prior density is specified with a sparse precision matrix. In the case of a large number of independent data replicates, sparse precision matrices, and high-dimensional latent vectors, the speedup is substantial in comparison to an MCMC scheme that infers the posterior density from the exact likelihood function. The accuracy of the Gaussian approximation to the likelihood function increases with the number of data replicates per latent model parameter. The proposed inference scheme is demonstrated on one spatially referenced real dataset and on simulated data mimicking spatial, temporal, and spatio-temporal inference problems. Our results show that Max-and-Smooth is accurate and fast.

Inference in Hybrid Bayesian Networks Using Mixtures of Gaussians

Bayesian estimation and prediction for certain mixtures

Gaussian Mixture Reduction for Time-Constrained Approximate Inference in Hybrid Bayesian Networks

Learning Gaussian Networks

Probabilistic Inference from Arbitrary Uncertainty using Mixtures of Factorized Generalized Gaussians

Continuity approximation in hybrid Bayesian networks structure learning

Continuous Gaussian mixture solution for linear Bayesian inversion with application to Laplace priors

Estimation of multiple networks in Gaussian mixture models.

Nearest Neighbor Dirichlet Mixtures

Inference and Learning in Hybrid Probabilistic Network.

Approximate inferences for Bayesian hierarchical generalised linear regression models

Pseudo Independent Conditional Approximation for Training the Mixtures of Gaussian Processes

Likelihood approximations via Gaussian approximate inference

Fast, accurate and lightweight sequential simulation-based inference using Gaussian locally linear mappings

Learning in a hybrid Bayesian network structure for causal analysis

Recognition Networks for Approximate Inference in BN20 Networks

Structure Learning for Hybrid Bayesian Networks

Max-and-Smooth: A Two-Step Approach for Approximate Bayesian Inference in Latent Gaussian Models

Approximate Bayesian inference for latent Gaussian models by using integrated nested Laplace approximations

Variational Bayes Approximations for Clustering via Mixtures of Normal Inverse Gaussian Distributions

Composite Inference for Gaussian Processes