Greedy Approaches to Symmetric Orthogonal Tensor Decomposition
Provides theoretical analysis of existing methods for tensor decomposition, which is relevant to signal processing, machine learning, and statistics, but the contribution is incremental.
The paper reviews and compares perturbation bounds for two greedy incremental rank-one approximation methods for symmetric orthogonal tensor decomposition, presenting numerical experiments and open questions.
Finding the symmetric and orthogonal decomposition (SOD) of a tensor is a recurring problem in signal processing, machine learning and statistics. In this paper, we review, establish and compare the perturbation bounds for two natural types of incremental rank-one approximation approaches. Numerical experiments and open questions are also presented and discussed.