Once a new centroid (representative point) is selected for each cluster, k-means reassigns every data point to the cluster represented by the centroid that is nearest to it. show that after reassigning clusters, the sum of s2 over all clusters is less than or equal to s2 before the reassignment