Asymmetric Weighted Cascade Model for Competitive Influence Maximization

Vipin Gunda,Archit Mehta

2024-10-24

Abstract:We introduce a modified Weighted Cascade model that integrates asymmetric budgets and product scores, providing new insights into the Generalized Asymmetric Influence Maximization problem, which we establish as NP-hard. Our simulations demonstrate that players with higher budgets possess a distinct advantage in networks characterized by larger diameters, whereas players with superior product scores exhibit a significant advantage in networks with smaller diameters. Moreover, we identify a robust linear relationship between graph size and the magnitude of influenced nodes. In densely connected networks we derive bounds for the probabilities of influence that are independent of network size. Our examination of Nash equilibria in this domain underscores the absence of a guaranteed pure Nash equilibrium, suggesting that the strategic enhancement of budgets or product scores may yield more substantial benefits than the pursuit of an optimal strategy in this context.

Social and Information Networks

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the **Competitive Influence Maximization problem**, especially when asymmetric budgets and product ratings are introduced. Specifically, the paper mainly focuses on the following aspects:

1. **Asymmetric competitive environment**: In the real world, different entities (such as companies) may have different initial resources (budgets) and product qualities (product ratings). This asymmetry makes the traditional Competitive Influence Maximization model not accurate enough to truly reflect the actual situation.

2. **NP - hard property**: The author proves that the generalized asymmetric Competitive Influence Maximization problem is NP - hard, which means that there is no general polynomial - time algorithm to solve this type of problem. This conclusion is obtained through the reduction proof of the Vertex Cover problem.

3. **Influence of network structure**: Through simulation experiments, the author explores the influence of different network structures (such as ring graphs, balanced binary trees, and dense networks) on the competition results. The study finds that in the case of a large network diameter, players with higher budgets have a significant advantage; while in the case of a small network diameter, players with higher product ratings are more advantageous.

4. **Existence of Nash equilibrium**: The author also studies whether there is a Pure Nash Equilibrium in the asymmetric Competitive Influence Maximization problem. Through specific simulation experiments, the author finds that there is no guaranteed Pure Nash Equilibrium, but there is a Mixed Nash Equilibrium.

### Specific problem description

- **Asymmetric Weighted Cascade Model**: This model takes into account the initial budget and product rating of each company in the network and defines the process of influence propagation on this basis. The model uses discrete random variables to determine whether a node is influenced by a certain company and which company it is influenced by.

- **Relationship between network structure and influence propagation**: The author verifies the influence of network structure (such as diameter and node degree) on the speed and range of influence propagation through simulation experiments. The research shows that when the network diameter is large, players with higher budgets are more dominant; while when the network diameter is small, players with higher product ratings are more dominant.

- **Lack of Nash equilibrium**: The author shows through simulation experiments that in the asymmetric Competitive Influence Maximization problem, there is no guaranteed Pure Nash Equilibrium. This suggests that players should focus more on increasing budgets or improving product ratings rather than looking for optimal strategies.

In summary, this paper aims to improve the existing Competitive Influence Maximization model by introducing asymmetric budgets and product ratings, and to deeply explore its performance under different network structures and its theoretical properties.