FKML0: A Matlab Routine for Sparse Fuzzy Clustering

A novel Matlab routine, called FKML0, is proposed to implement a variant of the standard Fuzzy K-Means algorithm. Such a variant allows for sparsity in the fuzzy membership degree matrix by adding a suitable regularization term based on the L0 norm to the Fuzzy K-Means loss function. The new clustering algorithm, called Fuzzy K-Means with L0 regularization, is able to assign, on the one hand, objects clearly belonging to a cluster with membership degrees equal to 1 (and 0 elsewhere) and, on the other hand, objects with unclear assignment with fuzzy membership degrees in the interval [0,1].