Partially exclusive condition for sequential fuzzy co-cluster extraction
Katsuhiro Honda, Akira Notsu, Hidetomo Ichihashi · 2011
Sequential fuzzy co-cluster extraction has been proven to be useful for collaborative filtering tasks by extracting user-item co-clusters, in which promising items are connected to the corresponding users in each co-cluster. Since some popular items can be shared by multiple clusters in collaborative filtering problems, exclusive conditions, which force objects to belong to only one cluster, were used only for users. In this paper, it is demonstrated that such user-only exclusive conditions may cause poor clustering results, and partially exclusive conditions, in which part of items are also forced to be exclusive, are introduced. In addition, connections with neural approaches for co-clustering are discussed, in which co-clustering problems are identified using a modified model of principal component analysis.