PARAFAC-based Multilinear Subspace Clustering for Tensor data

Subspace clustering has become an increasingly popular data analysis and machine learning method, whose main assumption is that data are generated from a union of linear subspaces modeling. While successful in many applications these methods do not take into account the multilinear structure of data such as images and video. Prompted by this observation, the present work introduces a multilinear subspace clustering scheme that exploits the structure of the data, and its performance is evaluated on synthetic and real datasets against state-of-the-art subspace clustering algorithms.