Abstract:Diffusion of information in networks is at the core of many problems in AI. Common examples include the spread of ideas and rumors as well as marketing campaigns. Typically, information diffuses at a non-linear rate, for example, if markets become saturated or if users of social networks reinforce each other's opinions. Despite these characteristics, this area has seen little research, compared to the vast amount of results for linear models, which exhibit less complex dynamics. Especially, when considering the possibility of re-infection, no fully rigorous guarantees exist so far.

What problem does this paper attempt to address?

This paper discusses the non-linear model of information diffusion in social networks, particularly in the presence of market saturation and social reinforcement. Traditional information diffusion models often assume linear propagation speed, but experiments have shown that some processes require consideration of non-linear effects, such as when the number of users increases, the infection rate (such as the adoption of opinions or product purchases) may increase or decrease. The paper proposes a non-linear diffusion model, where the probability of infection of each susceptible node is proportional to some exponent of the number of infected individuals in its neighborhood. The authors conducted a rigorous mathematical analysis of this non-linear SIS model for the first time, focusing on two cases: one where the infection rate grows at a sub-linear rate with the number of neighbors, and the other where it grows at a super-linear rate. They provided threshold values for the infection coefficient on complete graphs and star graphs, which differentiate between logarithmic and super-polynomial expected survival times. The results show that, on star graphs, regardless of whether the infection rate is sub-linear or super-linear, the threshold differs from that of the linear model by a polynomial factor. On complete graphs, the threshold for sub-linear infection rates is close to that of the linear model, while super-linear infection rates significantly lower the threshold and result in high variance in survival time. Overall, this paper rigorously analyzes how non-linear infection rates alter the behavior of the information diffusion model, particularly in terms of survival time and infection threshold in networks. These findings are significant for understanding information propagation phenomena in complex networks.

From Market Saturation to Social Reinforcement: Understanding the Impact of Non-Linearity in Information Diffusion Models

A Survey on Information Diffusion Models in Social Networks

The Irrelevance of Influencers: Information Diffusion with Re-Activation and Immunity Lasts Exponentially Long on Social Network Models

Information diffusion epidemics in social networks

Information Spreading Dynamics on Adaptive Social Networks

Evolutionary Dynamics of Information Diffusion over Social Networks

Evolutionary Information Diffusion over Social Networks

Modeling Information Diffusion in Social Networks

Reemergence Modeling of Intelligent Information Diffusion in Heterogeneous Social Networks: the Dynamics Perspective

Dynamics of Information Diffusion and Its Applications on Complex Networks

How Events Determine Spreading Patterns: Information Transmission Via Internal and External Influences on Social Networks

Information Spreading on Dynamic Social Networks

Characterizing Information Diffusion in Online Social Networks with Linear Diffusive Model

Information Diffusion in Social Networks: Friendship Paradox based Models and Statistical Inference

Evolutionary social information diffusion analysis

Inferring Networks of Diffusion and Influence

Marketing in Random Networks

Spreading in Online Social Networks: the Role of Social Reinforcement.

Understanding Correlated Information Diffusion: From a Graphical Evolutionary Game Perspective

Information diffusion through online social networks

Predictive Analysis for Social Diffusion: The Role of Network Communities