S. Wang et Sw. Xia, A SELF-ORGANIZING NETWORK FOR HYPERELLIPSOIDAL CLUSTERING (HEC) - COMMENTS, IEEE transactions on neural networks, 8(6), 1997, pp. 1561-1562
In the above paper,(1) Mahalanobis distance is used instead of Euclide
an distance as the distance measure in order to acquire the hyperellip
soidal clustering, We prove that the clustering cost function is a con
stant under this condition, so hyperellipsoidal clustering cannot be r
ealized, This letter also explains why the clustering algorithm develo
ped in the above paper can get some good hyperellipsoidal clustering r
esults.