Small-Coupling Dynamic Cavity: a Bayesian mean-field framework for epidemic inference

Abstract

We present the Small-Coupling Dynamic Cavity (SCDC) method, a novel generalized mean-field approximation for epidemic inference and risk assessment within a fully Bayesian framework. SCDC accounts for non-causal effects of observations and uses a graphical model representation of epidemic processes to derive self-consistent equations for edge probability marginals. A small-coupling expansion yields time-dependent cavity messages capturing individual infection probabilities and observational conditioning. With linear computational cost per iteration in the epidemic duration, SCDC is particularly efficient and valid even for recurrent epidemic processes, where standard methods are exponentially complex. Tested on synthetic networks, it matches Belief Propagation in accuracy and outperforms individual-based mean-field methods. Notably, despite being derived as a small-infectiousness expansion, SCDC maintains good accuracy even for relatively large infection probabilities. While convergence issues may arise on graphs with long-range correlations, SCDC reliably estimates risk. Future extensions include non-Markovian models and higher-order terms in the dynamic cavity framework.

28 pages, 11 figures, 2 tables (including appendices)

References in corpus (27)

• Epidemic processes in complex networks
• Discrete-time Markov chain approach to contact-based disease spreading in complex networks
• Inferring the origin of an epidemic with a dynamic message-passing algorithm
• Can co-location be used as a proxy for face-to-face contacts?
• Bayesian inference of epidemics on networks via Belief Propagation
• Identification of Patient Zero in Static and Temporal Networks - Robustness and Limitations
• Connectivity of soft random geometric graphs
• Optimal Deployment of Resources for Maximizing Impact in Spreading Processes
• Optimizing spread dynamics on graphs by message passing
• Dynamic message-passing equations for models with unidirectional dynamics
• Large deviations of cascade processes on graphs
• Majority dynamics on trees and the dynamic cavity method
• The Cavity Approach to Parallel Dynamics of Ising Spins on a Graph
• High-temperature Expansions and Message Passing Algorithms
• Dynamic mean-field and cavity methods for diluted Ising systems
• The zero-patient problem with noisy observations
• Epidemic mitigation by statistical inference from contact tracing data
• A message-passing scheme for non-equilibrium stationary states
• A Bayesian generative neural network framework for epidemic inference problems
• A Cavity Master Equation for the continuous time dynamics of discrete spins models
• Extended Plefka Expansion for Stochastic Dynamics
• A simple analytical description of the non-stationary dynamics in Ising spin systems
• Matrix Product Belief Propagation for reweighted stochastic dynamics over graphs
• Bayes-optimal inference for spreading processes on random networks
• Predicting epidemic evolution on contact networks from partial observations
• Inference in conditioned dynamics through causality restoration
• Effectiveness of probabilistic contact tracing in epidemic containment: the role of super-spreaders and transmission path reconstruction

Cited by in corpus (2)

• Evidence of Replica Symmetry Breaking under the Nishimori conditions in epidemic inference on graphs
• Nonequilibrium steady-state dynamics of Markov processes on graphs