Eigenvalue distribution of bipartite large weighted random graphs. Resolvent approach (1507.07529v1)

Abstract: We study eigenvalue distribution of the adjacency matrix $A^{(N,p, \alpha)}$ of weighted random bipartite graphs $\Gamma= \Gamma_{N,p}$. We assume that the graphs have $N$ vertices, the ratio of parts is $\frac{\alpha}{1-\alpha}$ and the average number of edges attached to one vertex is $\alpha\cdot p$ or $(1-\alpha)\cdot p$. To each edge of the graph $e_{ij}$ we assign a weight given by a random variable $a_{ij}$ with the finite second moment. We consider the resolvents $G^{(N,p, \alpha)}(z)$ of $A^{(N,p, \alpha)}$ and study the functions $f_{1,N}(u,z)=\frac{1}{[\alpha N]}\sum_{k=1}^{[\alpha N]}e^{-ua_k^2G_{kk}^{(N,p,\alpha)}(z)}$ and $f_{2,N}(u,z)=\frac{1}{N-[\alpha N]}\sum_{k=[\alpha N]+1}^Ne^{-ua_k^2G_{kk}^{(N,p,\alpha)}(z)}$ in the limit $N\to \infty$. We derive closed system of equations that uniquely determine the limiting functions $f_{1}(u,z)$ and $f_{2}(u,z)$. This system of equations allow us to prove the existence of the limiting measure $\sigma_{p, \alpha}$ . The weak convergence in probability of normalized eigenvalue counting measures is proved.