An Efficient Implementation of the Robust k-Center Clustering Problem

The standard k-center clustering problem is very sensitive to outliers. Charikar et al. proposed an alternative algorithm to cluster p points out of n total, thereby avoiding the distortion caused by outliers. The algorithm has an approximation bound of three times the true solution, but is very slow if implemented naively. We propose a modified implementation of the algorithm that runs significantly faster than the standard version. It does this while keeping the memory bound within the same asymptotic bound as that of the naive implementation. We show that, as the size of the problem increases, our algorithm maintains a relatively low running time, while the standard implementation time increases in proportion to