Random graphs are mathematical models that simulate complex networks. The Erdős-Rényi model, introduced in 1959, is a foundational approach for generating these graphs with specific properties.
This model comes in two variants: G(n,p) and G(n,M). Both create graphs with n vertices, but differ in how edges are assigned. Understanding these models is crucial for analyzing real-world networks in various fields.
Erdős-Rényi Random Graph Model Fundamentals
Erdős-Rényi random graph model
- Erdős-Rényi model generates random graphs with specific properties
- Named after mathematicians Paul Erdős and Alfréd Rényi who introduced it in 1959
- Two main variants G(n,p) and G(n,M) differ in how edges are assigned
- G(n,p) model uses n vertices and probability p for edge formation
- G(n,M) model uses n vertices and fixed number M of edges
- Widely used in network science to study complex systems (social networks, biological networks)
Generation of random graphs
- G(n,p) model generation process creates graphs with varying edge counts:
- Begin with n isolated vertices
- Examine each vertex pair
- Add edge between pair with probability p
- Repeat for all $\binom{n}{2}$ possible edges
- G(n,M) model generation process ensures exact edge count:
- Start with n isolated vertices
- Randomly choose M pairs of vertices
- Connect each selected pair with an edge
- G(n,p) uses probabilistic approach while G(n,M) guarantees fixed edge count
- Both models produce graphs with uniform distribution within their constraints
Probability of graph structures
- Probability of specific graph G with m edges in G(n,p) model calculated as $P(G) = p^m (1-p)^{\binom{n}{2} - m}$
- Expected number of edges in G(n,p) model given by $E(|E|) = \binom{n}{2}p$
- Connectivity probability increases with n and p, sharp threshold at $p = \frac{\ln n}{n}$
- Giant component emerges when $np > 1$, crucial for network connectivity
- k-clique formation probability: $\binom{n}{k}p^{\binom{k}{2}}$, important for community detection
- Threshold phenomena occur in both models as n approaches infinity
G(n,p) vs G(n,M) model properties
- Both produce n-vertex random graphs with uniform distribution
- G(n,p) allows variable edge count, G(n,M) fixes edge count at M
- G(n,M) equivalent to G(n,p) conditioned on exactly M edges
- Similar asymptotic behavior as n approaches infinity
- G(n,p) often easier for theoretical analysis
- G(n,M) useful when specific edge count required (sparse graph analysis)
- Computational trade-offs: G(n,p) faster for large n, G(n,M) more controlled