CANDECOMP/PARAFAC Decomposition of High-order Tensors Through Tensor Reshaping

Abstract

In general, algorithms for order-3 CANDECOMP/-PARAFAC (CP), also coined canonical polyadic decomposition (CPD), are easily to implement and can be extended to higher order CPD. Unfortunately, the algorithms become computationally demanding, and they are often not applicable to higher order and relatively large scale tensors. In this paper, by exploiting the uniqueness of CPD and the relation of a tensor in Kruskal form and its unfolded tensor, we propose a fast approach to deal with this problem. Instead of directly factorizing the high order data tensor, the method decomposes an unfolded tensor with lower order, e.g., order-3 tensor. On basis of the order-3 estimated tensor, a structured Kruskal tensor of the same dimension as the data tensor is then generated, and decomposed to find the final solution using fast algorithms for the structured CPD. In addition, strategies to unfold tensors are suggested and practically verified in the paper.

References in corpus (2)

• Cramer-Rao-Induced Bounds for CANDECOMP/PARAFAC tensor decomposition
• On Fast Computation of Gradients for CANDECOMP/PARAFAC Algorithms

Cited by in corpus (9)

• A Practical Randomized CP Tensor Decomposition
• DOA Estimation for Transmit Beamspace MIMO Radar via Tensor Decomposition with Vandermonde Factor Matrix
• Effective criteria for specific identifiability of tensors and forms
• Cramer-Rao-Induced Bounds for CANDECOMP/PARAFAC tensor decomposition
• Parallel numerical tensor methods for high-dimensional PDEs
• Mode-wise Tensor Decompositions: Multi-dimensional Generalizations of CUR Decompositions
• 5G Positioning and Mapping with Diffuse Multipath
• Two-Dimensional DOA Estimation for L-shaped Nested Array via Tensor Modeling
• Higher-Order Block Term Decomposition for Spatially Folded fMRI Data