Abstract:In multivariate statistics, the question of finding direct interactions can be formulated as a problem of network inference - or network reconstruction - for which the Gaussian graphical model (GGM) provides a canonical framework. Unfortunately, the Gaussian assumption does not apply to count data which are encountered in domains such as genomics, social sciences or ecology. To circumvent this limitation, state-of-the-art approaches use two-step strategies that first transform counts to pseudo Gaussian observations and then apply a (partial) correlation-based approach from the abundant literature of GGM inference. We adopt a different stance by relying on a latent model where we directly model counts by means of Poisson distributions that are conditional to latent (hidden) Gaussian correlated variables. In this multivariate Poisson lognormal-model, the dependency structure is completely captured by the latent layer. This parametric model enables to account for the effects of covariates on the counts. To perform network inference, we add a sparsity inducing constraint on the inverse covariance matrix of the latent Gaussian vector. Unlike the usual Gaussian setting, the penalized likelihood is generally not tractable, and we resort instead to a variational approach for approximate likelihood maximization. The corresponding optimization problem is solved by alternating a gradient ascent on the variational parameters and a graphical-Lasso step on the covariance matrix. We show that our approach is highly competitive with the existing methods on simulation inspired from microbiological data. We then illustrate on three various data sets how accounting for sampling efforts via offsets and integrating external covariates (which is mostly never done in the existing literature) drastically changes the topology of the inferred network.

Global multivariate model learning from hierarchically correlated data

A statistical physics approach to learning curves for the Inverse Ising problem

Learning Theory for Inferring Interaction Kernels in Second-Order Interacting Agent Systems

Inverse statistical problems: from the inverse Ising problem to data science

Spatial Factor Modeling: A Bayesian Matrix-Normal Approach for Misaligned Data

Joint Learning of Full-structure Noise in Hierarchical Bayesian Regression Models

Physics-aware, probabilistic model order reduction with guaranteed stability

Learning Interacting Theories from Data

Asymptotic theory with hierarchical autocorrelation: Ornstein-Uhlenbeck tree models

Inferring effective couplings with Restricted Boltzmann Machines

On the integration of Physics-Based Machine Learning with hierarchical Bayesian modeling techniques

Modeling massive highly-multivariate nonstationary spatial data with the basis graphical lasso

Variational inference for sparse network reconstruction from count data

High-dimensional learning of narrow neural networks

Online Variational Bayes Inference for High-Dimensional Correlated Data

Leveraging joint sparsity in hierarchical Bayesian learning

Physically optimizing inference

Scalable Physics-based Maximum Likelihood Estimation using Hierarchical Matrices

Learning maximum entropy models from finite-size data sets: A fast data-driven algorithm allows sampling from the posterior distribution

Hierarchical Neural Simulation-Based Inference Over Event Ensembles

Variational Hierarchical Mixtures for Probabilistic Learning of Inverse Dynamics