Propagação epidêmica em multigrafos
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidade Federal de Viçosa
Abstract
Multigrafos são redes que possuem múltiplas conexões entre vértices. Redes com distri- buição de grau em lei de potência P (k) ∼ k −γ apresentam um corte natural kN ∼ N γ−1 . Problemas de conectividade podem surgir para redes heterogêneas que permitem apenas uma conexão única entre pares, ou seja, grafos simples. Por um lado, o algoritmo pode não completar todas as conexões, resultando em alto custo computacional. Por outro lado, permitir múltiplas conexões permite gerar redes sem tais problemas. Portanto, motivado por uma perspectiva metodológica, este estudo propõe a propagação de epidemias em redes simples e multigrafos com kmax ∼ kN . Realizou-se análise quase estacionária (QS) para medir o limiar epidêmico, a densidade de indivı́duos infectados, a suscetibilidade dinâmica e a razão de participação inversa (RPI) usando o Algoritmo de Gillespie Otimizado (OGA). Assim, foram realizadas análises de tamanho finito a fim de comparar os expoentes de escala para o modelos Suscetı́vel-Infectado-Suscetı́vel (SIS), Suscetı́vel-Infectado-Removido- Suscetı́vel (SIRS) e Processo de Contato (PC) em grafos simples e multigrafos. Além disso, comparou-se o limiar epidêmico e a RPI previstos por teorias de campo médio com aqueles obtidos nas simulações estocásticas. Observou-se que o PC é invariante à presença de múltiplas conexões e que as previsões das teorias de campo médio concordam com os resultados obtidos nas simulações. O modelo SIRS não apresentou diferença no comportamento de escala para as grandezas mensuradas e os resultados concordam para simulações e campo médio para ambos tipos de grafos. Finalmente, as grandezas QS não mostraram diferença significativa no comportamento de escala para o modelo SIS com γ > 5/2. Para γ < 5/2, as grandezas apresentam diferenças no comportamento de escala, mas a fı́sica é preservada no limite termodinâmico. A implementação de multigrafos para a investigação de processos epidêmicos se mostrou adequada uma vez que não altera a fı́sica dos modelos epidêmicos nem a fı́sica das abordagens de campo médio. Palavras-chave: redes complexas, fenômenos crı́ticos, teorias de campo médio, fenômenos de localização.
Multigraphs are networks with multiple connections between vertices. Networks with a power-law degree distribution P (k) ∼ k −γ exhibit a natural cutoff kN ∼ N γ−1 . Connecti- vity issues can arise for heterogeneous networks that allow only a single connection between pairs, i.e., simple graphs. On one hand, the algorithm may fail to complete all connections, resulting in high computational cost. On the other hand, allowing multiple connections enables the generation of networks without these drawbacks. Therefore, motivated by a methodological perspective, this study investigates epidemic spreading on both simple networks and multigraphs with kmax ∼ kN . A quasi-stationary (QS) analysis was performed to measure the epidemic threshold, the density of infected individuals, dynamic susceptibi- lity, and the inverse participation ratio (IPR) using the Optimized Gillespie Algorithm (OGA). Additionally, finite-size analyses were conducted to compare the scaling exponents for the Susceptible-Infected-Susceptible (SIS), Susceptible-Infected-Removed-Susceptible (SIRS), and Contact Process (CP) models on simple graphs and multigraphs. Furthermore, the epidemic threshold and inverse participation ratio predicted by mean-field theories were compared with those obtained from stochastic simulations. It was observed that the CP is invariant to the presence of multiple connections, and that the predictions of mean-field theories agree with the results obtained from the simulations. The SIRS model showed no difference in scaling behavior for the measured quantities, and the results agreed for both simulations and mean-field theory for both types of graphs. Finally, the QS quantities showed no significant difference in scaling behavior for the SIS model with γ > 5/2. For γ < 5/2, the quantities exhibit differences in scaling behavior, but the physics is preserved in the thermodynamic limit. The implementation of multigraphs for investigating epidemic processes proved to be feasible as it does not alter the physics of the epidemic models nor the physics of mean-field approaches. Keywords: complex networks, critical phenomena, mean-field theories, localization phenomena.
Multigraphs are networks with multiple connections between vertices. Networks with a power-law degree distribution P (k) ∼ k −γ exhibit a natural cutoff kN ∼ N γ−1 . Connecti- vity issues can arise for heterogeneous networks that allow only a single connection between pairs, i.e., simple graphs. On one hand, the algorithm may fail to complete all connections, resulting in high computational cost. On the other hand, allowing multiple connections enables the generation of networks without these drawbacks. Therefore, motivated by a methodological perspective, this study investigates epidemic spreading on both simple networks and multigraphs with kmax ∼ kN . A quasi-stationary (QS) analysis was performed to measure the epidemic threshold, the density of infected individuals, dynamic susceptibi- lity, and the inverse participation ratio (IPR) using the Optimized Gillespie Algorithm (OGA). Additionally, finite-size analyses were conducted to compare the scaling exponents for the Susceptible-Infected-Susceptible (SIS), Susceptible-Infected-Removed-Susceptible (SIRS), and Contact Process (CP) models on simple graphs and multigraphs. Furthermore, the epidemic threshold and inverse participation ratio predicted by mean-field theories were compared with those obtained from stochastic simulations. It was observed that the CP is invariant to the presence of multiple connections, and that the predictions of mean-field theories agree with the results obtained from the simulations. The SIRS model showed no difference in scaling behavior for the measured quantities, and the results agreed for both simulations and mean-field theory for both types of graphs. Finally, the QS quantities showed no significant difference in scaling behavior for the SIS model with γ > 5/2. For γ < 5/2, the quantities exhibit differences in scaling behavior, but the physics is preserved in the thermodynamic limit. The implementation of multigraphs for investigating epidemic processes proved to be feasible as it does not alter the physics of the epidemic models nor the physics of mean-field approaches. Keywords: complex networks, critical phenomena, mean-field theories, localization phenomena.
Description
Citation
LORENZONI FILHO, Paulo Henrique. Propagação epidêmica em multigrafos. 2025. 59 f. Dissertação (Mestrado em Física) - Universidade Federal de Viçosa, Viçosa. 2025.
