c-means clustering with the l/sub l/ and l/sub infinity / norms

摘要: An extension of the hard and fuzzy c-means (HCM/FCM) clustering algorithms is described. Specifically, these models are extended to admit case where (dis)similarity measure on pairs numerical vectors includes two members Minkowski or p-norm family, viz., p=1 p= infinity norms. In absence theoretically necessary conditions guide a solution nonlinear constrained optimization problem associated with this case, it shown that certain basis exchange algorithm can be used find approximate critical points new objective functions. This method broadens applications horizon FCM family by enabling users match discontinuous multidimensional data structures similarity measures have nonhyperelliptical topologies. >