Tensor Factorization Based Method for Tensor Completion with Spatio-Temporal Characterization

Abstract

In this paper, we propose a novel tensor factorization based method for the third order tensor completion problem with spatio-temporal characterization. For this aim, we consider tensor fibered rank, which extends tubal rank, to improve the flexibility and accuracy of data characterization. Based on this rank, we apply a factorization based method to complete the third order low rank tensors with spatio-temporal characteristics, which are intrinsic features of image, video and internet traffic tensor data. The model not only makes good use of the low rank structure of tensors, but also takes into account the spatio-temporal characteristics of the data. Finally, we report numerical results on completing image, video and internet traffic data. The results demonstrate that our method outperforms some existing methods.

Quan Yu

PhD student

My research interests include low rank tensor optimization, image processing and machine learning.