Improvements on the k-center problem for uncertain data

In real applications, there are situations where we need to model some problems based on uncertain data. This leads us to define an uncertain model for some classical geometric optimization problems and propose algorithms to solve them. In this paper, we study the $k$-center problem, for uncertain input. In our setting, each uncertain point $Pi$ is located independently from other points in one of several possible locations ${P{i,1},\dots, P{i,zi}}$ in a metric space with metric $d$, with specified probabilities and the goal is to compute $k$-centers ${c1,\dots, ck}$ that minimize the following expected cost $$Ecost(c1,\dots, ck)=\sum{R\in \Omega} prob(R)\max{i=1,\dots, n}\min{j=1,\dots k} d(\hat{P}i,cj)$$ here $\Omega$ is the probability space of all realizations $$R={\hat{P}1,\dots, \hat{P}n}$$ of given uncertain points and $$prob(R)=\prod{i=1}ⁿ prob(\hat{P}i).$$ In restricted assigned version of this problem, an assignment $A:{P1,\dots, Pn}\rightarrow {c1,\dots, ck}$ is given for any choice of centers and the goal is to minimize $$EcostA(c1,\dots, ck)=\sum{R\in \Omega} prob(R)\max{i=1,\dots, n} d(\hat{P}i,A(Pi)).$$ In unrestricted version, the assignment is not specified and the goal is to compute $k$ centers ${c1,\dots, ck}$ and an assignment $A$ that minimize the above expected cost. We give several improved constant approximation factor algorithms for the assigned versions of this problem in a Euclidean space and in a general metric space. Our results significantly improve the results of \cite{guh} and generalize the results of \cite{wang} to any dimension. Our approach is to replace a certain center point for each uncertain point and study the properties of these certain points. The proposed algorithms are efficient and simple to implement.