Comparison of tolerant fuzzy c-means clustering with L2- and L1-regularization

Abstract

In this paper, we will propose two types of tolerant fuzzy c-means clustering with regularization terms. One is L2-regularization term and the other is L1-regularization one for tolerance vector. Introducing a concept of clusterwise tolerance, we have proposed tolerant fuzzy c-means clustering from the viewpoint of handling data more flexibly. In tolerant fuzzy c-means clustering, a constraint for tolerance vector which restricts the upper bound of tolerance vector is used. In this paper, regularization terms for tolerance vector are used instead of the constraint. First, the concept of clusterwise tolerance is introduced. Second, optimization problems for tolerant fuzzy c-means clustering with regularization term are formulated. Third, optimal solutions of these optimization problems are derived. Fourth, new clustering algorithms are constructed based on the explicit optimal solutions. Finally, effectiveness of proposed algorithms is verified through numerical examples.