Small-coupling dynamic cavity: A Bayesian mean-field framework for epidemic inference DOI Creative Commons
Alfredo Braunstein, Giovanni Catania, Luca Dall’Asta

et al.

Physical Review Research, Journal Year: 2025, Volume and Issue: 7(2)

Published: April 25, 2025

A novel generalized mean-field approximation, called the small-coupling dynamic cavity (SCDC) method, for epidemic inference and risk assessment is presented. The method developed within a fully Bayesian framework accounts noncausal effects generated by presence of observations. It based on graphical model representation stochastic process utilizes equations to derive set self-consistent probability marginals defined edges contact graph. By performing expansion, pair time-dependent messages obtained, which capture individual infection conditioning power In its efficient formulation, computational cost per iteration SCDC algorithm linear in duration dynamics. While derived susceptible-infected (SI) model, it straightforwardly applicable many other Markovian processes, including recurrent ones. This complexity particularly advantageous where methods are typically exponentially complex exhibits high accuracy assessing risk, as demonstrated tests SI applied various classes synthetic networks, performs par with belief propagation techniques generally exceeds performance individual-based methods. Additionally, was models, showed interesting even relatively large values probability, highlighting versatility effectiveness challenging scenarios. Although convergence issues may arise due long-range correlations graphs, estimated marginal probabilities remain sufficiently accurate reliable estimation. Future work includes extending non-Markovian models investigating role second-order terms expansion observation-reweighted equations. Published American Physical Society 2025

Language: Английский

Identifying Influential Nodes Based on Evidence Theory in Complex Network DOI Creative Commons

Fu Tan,

Xiaolong Chen, Rui Chen

et al.

Entropy, Journal Year: 2025, Volume and Issue: 27(4), P. 406 - 406

Published: April 10, 2025

Influential node identification is an important and hot topic in the field of complex network science. Classical algorithms for identifying influential nodes are typically based on a single attribute or simple fusion few attributes. However, these methods perform poorly real networks with high complexity diversity. To address this issue, new method Dempster–Shafer (DS) evidence theory proposed paper, which improves efficiency through following three aspects. Firstly, quantifies uncertainty its basic belief assignment function combines from different information sources, enabling it to effectively handle uncertainty. Secondly, processes conflicting using Dempster’s rule combination, enhancing reliability decision-making. Lastly, networks, may come multiple dimensions, can integrate multidimensional information. verify effectiveness method, extensive experiments conducted real-world networks. The results show that, compared other algorithms, attacking identified by DS more likely lead disintegration network, indicates that effective key network. further validate algorithm, we use visibility graph algorithm convert GBP futures time series into then rank method. top-ranked correspond peaks troughs series, represents turning points price changes. By conducting in-depth analysis, investors uncover major events influence trends, once again confirming algorithm.

Language: Английский

Citations

0
Small-coupling dynamic cavity: A Bayesian mean-field framework for epidemic inference DOI Creative Commons
Alfredo Braunstein, Giovanni Catania, Luca Dall’Asta

et al.

Physical Review Research, Journal Year: 2025, Volume and Issue: 7(2)

Published: April 25, 2025

A novel generalized mean-field approximation, called the small-coupling dynamic cavity (SCDC) method, for epidemic inference and risk assessment is presented. The method developed within a fully Bayesian framework accounts noncausal effects generated by presence of observations. It based on graphical model representation stochastic process utilizes equations to derive set self-consistent probability marginals defined edges contact graph. By performing expansion, pair time-dependent messages obtained, which capture individual infection conditioning power In its efficient formulation, computational cost per iteration SCDC algorithm linear in duration dynamics. While derived susceptible-infected (SI) model, it straightforwardly applicable many other Markovian processes, including recurrent ones. This complexity particularly advantageous where methods are typically exponentially complex exhibits high accuracy assessing risk, as demonstrated tests SI applied various classes synthetic networks, performs par with belief propagation techniques generally exceeds performance individual-based methods. Additionally, was models, showed interesting even relatively large values probability, highlighting versatility effectiveness challenging scenarios. Although convergence issues may arise due long-range correlations graphs, estimated marginal probabilities remain sufficiently accurate reliable estimation. Future work includes extending non-Markovian models investigating role second-order terms expansion observation-reweighted equations. Published American Physical Society 2025

Language: Английский

Citations

0