Optimal bound for survival time of the SIRS process on star graphs
Phuc Lam, Oanh Nguyen, Iris Yang · 2024
Read Optimal bound for survival time of the SIRS process on star graphs by Phuc Lam, Oanh Nguyen, Iris Yang (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-21138.txt.
Description
Abstract . We analyze the Susceptible-Infected-Recovered-Susceptible (SIRS) process, a continuous-time Markov chain frequently employed in epidemiology to model the spread of infections on networks. In this framework, infections spread as infected vertices recover at rate 1, infect susceptible neighbors independently at rate λ \lambda , and recovered vertices become susceptible again at rate α \alpha .
Catalog record
| CREATOR | Phuc Lam, Oanh Nguyen, Iris Yang |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 109 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-21138
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Optimal bound for survival time of the SIRS process on star graphs Thanks: Nguyen is supported by NSF Grant DMS–2246575 and the Salomon grant. Phuc Lam, Oanh Nguyen, Iris Yang Abstract . We analyze the Susceptible-Infected-Recovered-Susceptible (SIRS) process, a continuous-time Markov chain frequently employed in epidemiology to model the spread of infections on networks. In this framework, infections spread as infected vertices recover at rate 1, infect susceptible neighbors independently at rate λ \lambda , and recovered vertices become susceptible again at rate α \alpha . This model presents a significantly greater analytical challenge compared to the SIS model, which has consequently inspired a much more extensive and rich body of mathematical literature for the latter. Understanding the survival time, the duration before the infection dies out completely, is a fundamental question in this context. On general graphs, survival time heavily depends on the infection’s persistence around high-degree vertices (known as hubs or stars), as long persistence enables transmission between hubs and prolongs the process. In contrast, short persistence leads to rapid extinction, making the dynamics on star graphs, which serve as key representatives of hubs, particularly important to study. In the 2016 paper by Ferreira, Sander, and Pastor-Satorras, it was conjectured, based on intuitive arguments, that the survival time for SIRS on a star graph with n n leaves is bounded above by ( λ 2 n ) α (\lambda^{2}n)^{\alpha} for large n n . Later, in one of a few mathematically rigorous results for SIRS, Friedrich, Göbel, Klodt, Krejca, and Pappik provided an upper bound of n α log n n^{\alpha}\log n , with contains an additional log n \log n and no dependence on λ \lambda . We resolve this conjecture by proving that the survival time is indeed of order ( λ 2 n ) α (\lambda^{2}n)^{\alpha} , with matching upper and lower bounds. Additionally, we show that this holds even in the case where only the root undergoes immunization, while the leaves revert to susceptibility immediately after recovery. 1. Introduction Our goal in this paper is to study the exact order of the survival time of the SIRS process on star graphs, a building block for future studies of the process on random networks. The SIRS contact process models the spread of disease on networks. For a graph G=(V,E) , the SIRS process on G with infection rate \lambda , recovery rate 1 , and deimmunization rate \alpha is a continuous-time Markov chain, in which a vertex is either susceptible, infected, or recovered. The process evolves according to the following rules. • Each infected vertex infects each of its neighbors indepedently at rate \lambda , and is healed at rate 1 . • Each recovered vertex is immune to infection before becoming susceptible again at rate \alpha . • Infection, recovery, and deimmunization events in the process happen independently. This model accounts for the temporal period of immunity, as observed in some diseases such as with influenza or COVID-19 and has been used to model the spread of epidemics and forest fires with numerous empirical results, such as [ 12 , 17 , 16 , 27 ] and the references therein. However, the mathematical study of the SIRS process is notoriously challenging, with only a few theoretical results available, including [ 5 ] for \mathbb{Z}^{2} and [ 8 ] . We also refer to Durrett’s recent book [ 7 ] (see Chapter 4) for a relevant discussion. A variant, more tractable version of this model, which can be thought of as a special case of this process when \alpha=\infty (no immunization), is known as the SIS proces s, introduced by Harris in 1974 [ 10 ] . Since then, the survival time of the SIS process has been extensively studied on a variety of graphs. On infinite structures, phase-type results for the infection rate \lambda has been established on the infinite integer lattice \mathbb{Z}^{d} [ 10 ] by Harris, infinite d -regular trees by Liggett [ 13 ] , Pemantle [ 23 ] , Stacey [ 25 ] , d -regular trees of depth h by Cranston et al. [ 2 ] , Stacey [ 26 ] ), the d -dimensional finite lattice cube by Durrett-Liu [ 4 ] , Durrett-Schonmann [ 6 ] , Mountford [ 18 ] ), random d -regular graphs by Lalley-Su [ 15 ] , Mourrat-Valesin [ 20 ] , Galton-Watson trees and sparse random graphs by Bhamidi-Nam-Sly and the second author [ 1 , 21 , 22 ] , and Huang-Durrett [ 11 ] , together with the references therein. In this paper, we study star graphs . To illustrate the importance of the study of contact process on star graphs , we note that from previous results for SIS, it has been observed that the survival time of the contact process inherits a significant influence by its survival time on local neighborhoods of high-degree vertices. For instance, in Bhamidi–Nam–Nguyen–Sly [ 1 ] and Huang–Durrett [ 11 ] , it was shown that the sufficient and necessary condition for the SIS on the Galton-Watson tree with offspring distribution \mu exhibits the extinction phase (i.e., dies out with probability 1) is that \mu does not have an exponential tail: there exists c>0 such that \mathbb{E}e^{cD}<\infty where D\sim\mu . This relates directly to the fact that the SIS survives for an exponentially long time on stars. Theorem 1.1 . [ 22 ] Consider the SIS process (X_{t}) with infection rate \lambda on the star graph with n leaves, where the root is infected at the beginning. Then when \lambda\geq\frac{10^{2.5}}{\sqrt{n}} , the following holds: with positive probability, the survival time of the SIS process is at least \exp\left(\frac{\lambda^{2}n}{10^{5}}\right)=\exp\left(\Theta(\lambda^{2}n)\right). Roughly speaking, if \mu has an exponential tail \mathbb{E}e^{cD}<\infty , then the process stays at the immediate neighborhood of a vertex of high degree D for time e^{c\lambda^{2}D} . If \lambda is sufficiently small compared to c , the time is not long enough for the contact process to try and reach another high-degree vertex. Going back to the SIRS, in the 2016 paper [ 9 , Equation 13] , Ferreira, Sander and Pastor-Satorras predicted an upper bound for the survival time of the SIRS on an n -star graph (i.e., a star graph with n leaves) to be (\lambda^{2}n)^{\alpha} . In [ 8 , Theorem 3.4] , Friedrich, Göbel, Klodt, Krejca and Pappik showed this upper bound, with an extra \log n term and no specific dependence on \lambda . (It was again conjectured in [ 8 ] that the \log n term is redundant.) Theorem 1.2 . [ 8 ] Let G be a star graph on n leaves, and C be the SIRS process on the graph G with infection rate \lambda and deimmunization rate \alpha . Let \tau be the survival time of C . Then for sufficiently large n , the expected survival time is at most \mathbb{E}\tau\leq(\log n+2)(4n^{\alpha}+1)=O(n^{\alpha}\log n). In this paper, we resolve this conjecture and determine the dependence on \lambda in the most interesting regime, namely when \lambda^{2}n\gg 1 . In the other regime, we show that the survival time is at most \log n . Theorem 1.3 . Let X=(X_{t})_{t\geq