Maximizing the size of the giant

Abstract

We consider two classes of random graphs: $(a)$ Poissonian random graphs in which the $n$ vertices in the graph have i.i.d.\ weights distributed as $X$, where $\mathbb{E}(X) = \mu$. Edges are added according to a product measure and the probability that a vertex of weight $x$ shares and edge with a vertex of weight $y$ is given by $1-e^{-xy/(\mu n)}$. $(b)$ A thinned configuration model in which we create a ground-graph in which the $n$ vertices have i.i.d.\ ground-degrees, distributed as $D$, with $\mathbb{E}(D) = \mu$. The graph of interest is obtained by deleting edges independently with probability $1-p$. In both models the fraction of vertices in the largest connected component converges in probability to a constant $1-q$, where $q$ depends on $X$ or $D$ and $p$. We investigate for which distributions $X$ and $D$ with given $\mu$ and $p$, $1-q$ is maximized. We show that in the class of Poissonian random graphs, $X$ should have all its mass at 0 and one other real, which can be explicitly determined. For the thinned configuration model $D$ should have all its mass at 0 and two subsequent positive integers.