Monotone Increasing Properties and Their Phase Transitions in Uniform Random Intersection Graphs

TL;DR

This paper analyzes phase transitions in uniform random intersection graphs, focusing on properties like perfect matching and Hamilton cycles, and computes the widths of these transitions under various parameter regimes.

Contribution

It provides exact probability analyses for key graph properties and characterizes phase transition widths across different growth conditions of parameters.

Findings

Graph properties exhibit phase transitions from 0 to 1 probability as $K_n$ increases.

Phase transition widths vary significantly depending on the growth rate of $P_n$.

For large $P_n$, transition widths for certain properties become unbounded.

Abstract

Uniform random intersection graphs have received much interest and been used in diverse applications. A uniform random intersection graph with $n$ nodes is constructed as follows: each node selects a set of $K_n$ different items uniformly at random from the same pool of $P_n$ distinct items, and two nodes establish an undirected edge in between if and only if they share at least one item. For such graph denoted by $G(n, K_n, P_n)$, we present the following results in this paper. First, we provide an exact analysis on the probabilities of $G(n, K_n, P_n)$ having a perfect matching and having a Hamilton cycle respectively, under $P_n = \omega\big(n (\ln n)^5\big)$ (all asymptotic notation are understood with $n \to \infty$). The analysis reveals that just like ($k$-)connectivity shown in prior work, for both properties of perfect matching containment and Hamilton cycle containment, $G(n, K_n, P_n)$ also exhibits phase transitions: for each property above, as $K_n$ increases, the limit of the probability that $G(n, K_n, P_n)$ has the property increases from $0$ to $1$. Second, we compute the phase transition widths of $G(n, K_n, P_n)$ for $k$-connectivity (KC), perfect matching containment (PMC), and Hamilton cycle containment (HCC), respectively. For a graph property $R$ and a positive constant $a < \frac{1}{2}$, with the phase transition width $d_n(R, a)$ defined as the difference between the minimal $K_n$ ensuring $G(n, K_n, P_n)$ having property $R$ with probability at least $1-a$ or $a$, we show for any positive constants $a<\frac{1}{2}$ and $k$: (i) If $P_n=\Omega(n)$ and $P_n=o(n\ln n)$, then $d_n(KC, a)$ is either $0$ or $1$ for each $n$ sufficiently large. (ii) If $P_n=\Theta(n\ln n)$, then $d_n(KC, a)=\Theta(1)$. (iii) If $P_n=\omega(n\ln n)$, then $d_n(KC, a)=\omega(1)$. (iv) If $P_n=\omega\big(n (\ln n)^5\big)$, $d_n(PMC, a)$ and $d_n(HCC, a)$ are both $\omega(1)$.