Low-rank sparse fully-connected tensor network for tensor completion

Jinshi Yu, Zhifu Li, Ge Ma, Jingwen Wang, Tao Zou, Guoxu Zhou

Abstract: Highlights•The existing FCTN method often has overfitting problem caused by rank-selection.•FCTN can be expressed by a coefficient sum of basic tensors.•Sparsity and low-rank constrain on FCTN’s factors can improve the rank-robustness.•A Low-rank sparse FCTN method for tensor completion is developed by ADMM.