Back to Videos


In this talk we discuss computational issues related to low-rank tensor completion and decomposition problems. Computing the so-called CP rank of a given tensor is already notoriously difficult, let alone the completion or decomposition models where the CP-rank plays the role of a regularization function. In this talk, we introduce various matricization schemes, and discuss their numerical performances. The talk is based on a joint work with Bo Jiang and Shiqian Ma.