Local equivalences of distances between clusterings -- a geometric perspective (Q420920): Difference between revisions

Many different similarity distances are used in comparing clustering methods. The authors try to prove geometrically, using the concavity of distances, that the three distances misclassification error, Hamming (equivalent to unadjusted Rand index), and the chi-square distance are equivalent in a neighborhood of zero. They claim that, to their knowledge, this is the first result on this subject.