25  Distance, Clustering, and Extremes

Builds k-means, then shows a single extreme feature value dragging a centroid away from the rest of its cluster, then builds k-medoids (minimizing summed absolute distance rather than squared distance) as the robust alternative.