Using the replica method, we derive quantitative information about the
clustering into 2 subgroups of data lying on the surface of an N-dime
nsional hypersphere. Various clustering algorithms are considered, inc
luding the one corresponding to competitive learning.