Abstract:We investigate the number of maximal cliques, i.e., cliques that are not contained in any larger clique, in three network models: Erdős-Rényi random graphs, inhomogeneous random graphs (also called Chung-Lu graphs), and geometric inhomogeneous random graphs. For sparse and not-too-dense Erdős-Rényi graphs, we give linear and polynomial upper bounds on the number of maximal cliques. For the dense regime, we give super-polynomial and even exponential lower bounds. Although (geometric) inhomogeneous random graphs are sparse, we give super-polynomial lower bounds for these models. This comes from the fact that these graphs have a power-law degree distribution, which leads to a dense subgraph in which we find many maximal cliques. These lower bounds seem to contradict previous empirical evidence that (geometric) inhomogeneous random graphs have only few maximal cliques. We resolve this contradiction by providing experiments indicating that, even for large networks, the linear lower-order terms dominate, before the super-polynomial asymptotic behavior kicks in only for networks of extreme size.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to study the number of maximum cliques (i.e., complete sub - graphs not contained in any larger clique) in different network models. Specifically, the author focuses on three network models: Erdős–Rényi random graphs, non - uniform random graphs (also known as Chung–Lu graphs), and geometric non - uniform random graphs. The main objective of the paper is to give upper and lower bounds on the number of maximum cliques in these models and explore how these bounds change as the number of network nodes \(n\) increases.
### Research Background and Motivation
1. **Characteristics of Real - World Networks**
- Many real - world networks have a heavy - tailed degree distribution (scale - free), which means that a few nodes have a large number of connections, while most nodes have only a small number of connections.
- These networks usually have a high clustering coefficient, that is, the neighbors of a node also tend to be connected to each other.
2. **Limitations of Existing Models**
- Although existing random graph models (such as hypergeometric random graphs and geometric non - uniform random graphs) can have both a heavy - tailed degree distribution and a high clustering coefficient, whether they match real - world data in other properties still needs further study.
3. **Importance of Maximum Cliques**
- Maximum cliques are an important indicator of network structure, reflecting the degree of aggregation of nodes in the network. Small cliques (such as triangles) can indicate the transitivity or clustering coefficient of the network.
- Listing all maximum cliques is crucial for understanding the structure of the network, because all smaller cliques can be generated from at least one maximum clique.
### Main Research Contents
1. **Erdős–Rényi Random Graphs**
- For sparse graphs (the number of edges \(O(n)\)), a linear upper bound is given.
- For non - dense graphs (the number of edges \(O(n^{2-\epsilon})\)), a polynomial upper bound is given.
- For dense graphs (the number of edges \(\Omega(n^{2})\)), a super - polynomial lower bound is given, and even when the density is high enough, an exponential lower bound is given.
2. **Non - uniform Random Graphs (Chung–Lu Graphs)**
- Although overall sparse, due to the existence of a power - law degree distribution, these graphs contain large enough dense sub - graphs, so that the lower bound of the number of maximum cliques is super - polynomial.
3. **Geometric Non - uniform Random Graphs (GIRG)**
- By constructing specific sub - graphs (such as co - matching graphs), it is proved that under certain conditions, the number of maximum cliques in GIRG can reach a super - polynomial lower bound.
- Specifically, for \(\tau\in(2, 3)\), the lower bound of the number of maximum cliques is \(\exp(\Omega(n^{(3 - \tau)/4-\epsilon}))\).
### Experimental Results
- The author verifies the theoretical results through experiments and finds that in actual networks, the growth rate of the number of maximum cliques is usually slow, and the linear term dominates in smaller - scale networks.
- Only when the network scale is very large does super - polynomial growth become significant.
### Conclusions
- This research reveals the influence of network density on the number of maximum cliques, especially the differences in sparse and dense graphs.
- Through theoretical analysis and experimental verification, the author resolves the contradiction between previous empirical observations and theoretical results, providing a deeper understanding of the behavior of the number of maximum cliques in real - world networks.
### Related Work
- This paper also discusses the comparison with other related work, especially compares the research results with hypergeometric random graphs (HRG) and random geometric graphs (RGG), and finds that although GIRG is a generalization of HRG, there are asymptotic differences in the number of maximum cliques.