The k-Center Problem of Uncertain Points on Graphs

In this paper, we study the $k$-center problem of uncertain points on a graph. Given are an undirected graph $G = (V, E)$ and a set $\mathcal{P}$ of $n$ uncertain points where each uncertain point with a non-negative weight has $m$ possible locations on $G$ each associated with a probability. The problem aims to find $k$ centers (points) on $G$ so as to minimize the maximum weighted expected distance of uncertain points to their expected closest centers. No previous work exist for the $k$-center problem of uncertain points on undirected graphs. We propose exact algorithms that solve respectively the case of $k=2$ in $O(|E|^2m^2n\log |E|mn\log mn )$ time and the problem with $k\geq 3$ in $O(\min\{|E|^km^kn^{k+1}k\log |E|mn\log m, |E|^kn^\frac{k}{2}m^\frac{k^2}{2}\log |E|mn\})$ time, provided with the distance matrix of $G$. In addition, an $O(|E|mn\log mn)$-time algorithmic approach is given for the one-center case.