Greedy Sampling for Approximate Clustering in the Presence of Outliers