Inference in spreading processes with neural-network priors
- 7 January 2026
- journal article
- Published by American Physical Society (APS) in Physical Review E
- Vol. 113 (1) , 015301
Abstract
Stochastic processes on graphs are a powerful tool for modeling complex dynamical systems such as epidemics. A recent line of work focused on the inference problem where one aims to estimate the state of every node at every time, starting from partial observation of a subset of nodes at a subset of times. In these works, the initial state of the process was assumed to be random independent and identically distributed over nodes. Such an assumption may not be realistic in practice, where one may have access to a set of covariate variables for every node that influences the initial state of the system. In this work, we will assume that the initial state of a node is an unknown function of such covariate variables. Given that functions can be represented by neural networks, we will study a model where the initial state is given by a simple neural network-notably the single-layer perceptron acting on the known nodewise covariate variables. Within a Bayesian framework, we study how such neural-network prior information enhances the recovery of initial states and spreading trajectories. We derive a hybrid belief propagation and approximate message passing (BP-AMP) algorithm that handles both the spreading dynamics and the information included in the node covariates, and we assess its performance against the estimators that use only the spreading information or use only the information from the covariate variables. We show that in some regimes, the model can exhibit first-order phase transitions when using a Rademacher distribution for the neural-network weights. These transitions create a statistical-to-computational gap where even the BP-AMP algorithm, despite the theoretical possibility of perfect recovery, fails to achieve it.Keywords
All Related Versions
Funding Information
- Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (212049)
- Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (200390)
This publication has 37 references indexed in Scilit:
- Equivalence of approximate message passing and low-degree polynomials in rank-one matrix estimationProbability Theory and Related Fields, 2024
- Modeling, estimation, and analysis of epidemics over networks: An overviewAnnual Reviews in Control, 2020
- Optimal errors and phase transitions in high-dimensional generalized linear modelsProceedings of the National Academy of Sciences, 2019
- Information-theoretic thresholds from the cavity methodAdvances in Mathematics, 2018
- Mathematics of Epidemics on Networks2017
- Modelling dynamical processes in complex socio-technical systemsNature Physics, 2011
- Modelling and analysis of gene regulatory networksNature Reviews Molecular Cell Biology, 2008
- Complex networks: Structure and dynamicsPhysics Reports, 2006
- Modern Graph Theory1998
- Storage capacity of memory networks with binary couplingsJournal de Physique, 1989