Clustering algorithms such as k$-$means and agglomerative hierarchical clustering can be applied

to a set of CF$-$values $($such as the CF$-$values in the leaf nodes of a CF$-$tree$)$ instead of the original

dataset. You have shown in a$)$ that the centroid, radius etc. can be computed exactly from the CFvalue without knowing the corresponding set of points. Nevertheless, the clustering of the CFvalues in the leaf nodes of a CF$-$tree can still be different from the clustering obtained when

applying the same algorithm to the original dataset. Explain why this can happen.