One-mode additive clustering of multiway data
Dirk Depril, Iven Van Mechelen · 2005
Abstract. Given a multiway data set, in several contexts it may be desirable to obtain an overlapping clustering of one of the modes implied by the data. For this purpose a one-mode additive clustering model has been proposed, which implies a decomposition of the data into a binary membership matrix and a real-valued centroid matrix. To fit this model to a data set, a least squares loss function can be minimized. This can be achieved by means of a sequential fitting algorithm (SEFIT) as developed by Mirkin. In this presentation we will propose a new algorithm for the same model, based on an alternating least squares approach.