An accelerated approach for low rank tensor completion

Tensor completion problem, which is a generalization of the matrix completion problem,is recovering the missing data of a tensor. In many algorithms proposed to complete thetensor, to achieve the answer, the method is executed on all unfolds related to tensor modes.Hence, if a tensor has N modes, each iteration of the algorithm contains N sub-problems,which is equivalent to solving N matrix completion problems. In this paper, to overcome thecomputational complexity caused by applying the algorithm to each dimension of the tensor,we present an idea in which the problem is implemented on only one tensor unfolding, so itscomputational complexity decreases in each iteration.

Keywords:

Tensor completion problem , n-rank , computational complexity.