Computation of the nonnegative canonical tensor decomposition with two accelerated proximal gradient algorithms

Multidimensional signal analysis has become an important part of many processing problems. This type allows to take advantage different diversities a in order extract useful information. paper focuses on the design and development multidimensional data decomposition algorithms called Canonical Polyadic (CP) tensor decomposition, powerful tool variety real-world applications due its uniqueness ease interpretation factor matrices. More precisely, it is desired compute simultaneously matrices involved CP real nonnegative tensor, under constraints. For this purpose, two proximal are proposed, Monotone Accelerated Proximal Gradient (M-APG) Non-monotone (Nm-APG) algorithms. These implemented via regularization function with simple control strategy capable efficiently taking previous iterations. Simulation results demonstrate better performance proposed terms accuracy when compared other literature.