Abstract:The generalised random graph contains $n$ vertices with positive i.i.d. weights. The probability of adding an edge between two vertices is increasing in their weights. We require the weight distribution to have finite second moments and study the point process $\mathcal{C}_n$ on $\{3,4,\dots\}$, which counts how many cycles of the respective length are present in the graph. We establish convergence of $\mathcal{C}_n$ to a Poisson point process. Under the stronger assumption of the weights having finite fourth moments we provide the following results. When $\mathcal{C}_n$ is evaluated on a bounded set $A$, we provide a rate of convergence. If the graph is additionally subcritical, we extend this to unbounded sets $A$ at the cost of a slower rate of convergence. From this we deduce the limiting distribution of the length of the shortest and the longest cycle when the graph is subcritical, including rates of convergence. All mentioned results also apply to the Chung-Lu model and the Norros-Reittu model.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: in the generalised random graph, study the convergence of the point process \(C_n\) on the set \(\{3, 4,\cdots\}\), where \(C_n\) counts the number of cycles of different lengths. Specifically, the author focuses on the following points:
1. **Poisson approximation**:
- When the weight distribution has a finite second moment, prove that \(C_n\) converges to a Poisson point process.
- Under stronger assumptions (that is, the weights have a finite fourth moment), provide the convergence rate when \(C_n\) is evaluated on a bounded set \(A\).
2. **Results in sub - critical graphs**:
- When the graph is sub - critical (that is, \(E[W^2]<E[W]\)), extend these results to an unbounded set \(A\), although the convergence rate is slower.
- Derive the limit distributions of the shortest and longest cycle lengths and their convergence rates.
3. **Applications of specific models**:
- Prove that the above results also apply to the Chung - Lu model and the Norros - Reittu model.
### Formula summary
- **Connection probability formula**:
\[
P_W(i\leftrightarrow j)=\frac{W_iW_j}{W_iW_j + L_n}\quad\text{where}\quad L_n=\sum_{i = 1}^nW_i
\]
- **Parameter definition**:
\[
\lambda_k=\frac{1}{2k}\left(\frac{E[W^2]}{E[W]}\right)^k
\]
- **Total variation distance formula**:
\[
d_{TV}(C_n(A),\eta(A))\leq\frac{C}{n}
\]
where \(C\) is a constant, \(A\subseteq\{3, 4,\cdots,N\}\).
- **Kolmogorov distance formula**:
\[
d_{Kol}(C^{(n)}_{\min},S)\leq\frac{C\log(n)^3}{n}
\]
\[
d_{Kol}(C^{(n)}_{\max},L)\leq\frac{C\log(n)^3}{n}
\]
### Main conclusions
1. **Theorem 2.1**: If \(E[W^2]<\infty\), then \(C_n\) converges in distribution to the Poisson point process \(\eta\).
2. **Theorem 2.2**: If \(E[W^4]<\infty\), then for all \(A\subseteq\{3, 4,\cdots,N\}\), there exists a constant \(C > 0\) such that
\[
d_{TV}(C_n(A),\eta(A))\leq\frac{C}{n}
\]
3. **Theorem 2.3**: If \(E[W^2]<E[W]\), then for any \(a > 0\), there exists a constant \(C > 0\) such that
\[
P(C_n(\lfloor a\log(n)\rfloor + 1,\cdots,n)>0)\leq Cn^{a\log(p)}
\]
where \(p=\frac{2E[W^2]}{E[W^2]+E[W]}\).
4. **Theorem 2.4**: If \(E[W^4]<\infty\) and \(E[W^2]<E[W]\), then there exists a constant \(C > 0\) such that for all \(A\subseteq\mathbb{N}\)