Ph. Blanchard,T. Krueger
Abstract:We formulate a simple edge generation rule based on an inverse like mass action principle for random graphs over a structured vertex set. We show that under very weak assumptions on the structure generating distribution we obtain a scale free distribution for the degree. We furthermore introduce and study a "my friends are your friends" local search principle which makes the clustering coefficient large.
What problem does this paper attempt to address?
### Problems the Paper Attempts to Solve
This paper attempts to explain the origin and formation mechanism of scale-free graphs in social networks. Specifically, the authors propose a simple edge generation rule based on the "anti-mass action principle" and demonstrate that a scale-free degree distribution can be obtained under very weak structural assumptions. Additionally, the paper introduces and studies a local search principle of "my friend is also your friend" to increase the clustering coefficient.
### Main Contributions
1. **Proposing the "Anti-Mass Action Principle"**:
- The authors assume that each vertex has a positive real parameter \(\omega\), representing its social importance, wealth, beauty, or other attributes.
- Edge generation is based on an inverse mass action principle, where the probability of selecting a vertex \(y\) is inversely proportional to the relative frequency of its parameter \(\omega(y)\).
- This principle captures the essence of many network formation processes in real life.
2. **Demonstrating the Emergence of Scale-Free Distribution**:
- Under the above assumptions, regardless of the distribution of \(\omega\), a scale-free degree distribution will eventually emerge.
- The authors prove that when the tail decay rate of \(\phi(\omega)\) exceeds any power law, the exponent of the degree distribution is independent of \(\phi(\omega)\).
3. **Introducing the "My Friend is Also Your Friend" Principle**:
- To increase the clustering coefficient, the authors introduce a local edge generation rule, which selects the neighbor's neighbor with a certain probability.
- This principle is very common in social networks and can significantly increase the network's clustering coefficient.
### Model Analysis
1. **Basic Assumptions**:
- The parameter \(\omega\) is independently and identically distributed over the vertex set, with a smooth monotonically decreasing density function \(\phi(\omega)\).
- Edge generation is a selection process where each vertex can choose other vertices to form edges.
- The selection probability is given by \( \text{Pr}\{x \rightarrow y | \omega(y)\} \sim [\phi(\omega(y))]^{-\alpha} \cdot \frac{1}{N} \), where \(\alpha \in (0, 1)\).
2. **Analysis of Degree Distribution**:
- The authors mathematically prove that under the above assumptions, the degree distribution exhibits scale-free characteristics.
- Specifically, the asymptotic form of the degree distribution is \( \psi(k) \sim \frac{1}{k^{1 + \frac{1}{\alpha}}} \).
3. **Analysis of Clustering Coefficient**:
- Without explicit local clustering rules, the model's clustering coefficient is slightly higher than that of the classic random graph model \(G(N, p)\).
- By introducing the "my friend is also your friend" principle, the clustering coefficient can be significantly increased.
### Conclusion and Outlook
- The authors demonstrate that a simple edge generation rule based on the "anti-mass action principle" can naturally generate scale-free graphs.
- By introducing local edge generation rules, the clustering coefficient can be further increased, making the network more consistent with the characteristics of real social networks.
- Future research directions include the support for different \(\phi(\omega)\) distributions, the model's performance under finite size effects, and comparisons with other network models.