When a data set has significant differences in its class and cluster
structure, selecting features aiming only at the discrimination of classes
would lead to poor clustering performance, and similarly, feature selection
aiming only at preserving cluster structures would lead to poor cl