Tensor ring decomposition: optimization landscape and one-loop convergence of alternating least squares

DOI10.1137/19M1270689MaRDI QIDQ5146616zbMATH OpenOpenAlexFDO

Authors Ziang Chen, Yingzhou Li, Jianfeng Lu

Publication date 26 January 2021

Published in SIAM Journal on Matrix Analysis and Applications (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1905.07101

zbMATH Keywords

alternating least squaresfirst iteration convergencespurious local minimatensor ring decomposition

Mathematics Subject Classification ID

Multilinear algebra, tensor calculus (15A69) Numerical linear algebra (65F99)

Abstract: In this work, we study the tensor ring decomposition and its associated numerical algorithms. We establish a sharp transition of algorithmic difficulty of the optimization problem as the bond dimension increases: On one hand, we show the existence of spurious local minima for the optimization landscape even when the tensor ring format is much over-parameterized, i.e., with bond dimension much larger than that of the true target tensor. On the other hand, when the bond dimension is further increased, we establish one-loop convergence for alternating least square algorithm for tensor ring decomposition. The theoretical results are complemented by numerical experiments for both local minimum and one-loop convergence for the alternating least square algorithm.

Recommendations

• Local convergence of the alternating least squares algorithm for canonical tensor approximation
• Some convergence results on the regularized alternating least-squares method for tensor decomposition
• On the global convergence of the alternating least squares method for rank-one approximation to generic tensors
• On local convergence of alternating schemes for optimization of convex problems in the tensor train format
• Optimization landscape of Tucker decomposition

Cites work

• A new scheme for the tensor representation
• A practical introduction to tensor networks: Matrix product states and projected entangled pair states
• An area law for one-dimensional quantum systems
• Breaking the Curse of Dimensionality, Or How to Use SVD in Many Dimensions
• CheMPS2: a free open-source spin-adapted implementation of the density matrix renormalization group for ab initio quantum chemistry
• Coordinatewise descent methods for leading eigenvalue problem
• Entanglement renormalization: an introduction
• Exponential decay of correlations implies area law
• Hierarchical Singular Value Decomposition of Tensors
• On local convergence of alternating schemes for optimization of convex problems in the tensor train format
• On the geometry of tensor network states
• Rigorous RG algorithms and area laws for low energy eigenstates in 1D
• Singular vectors of orthogonally decomposable tensors
• Spectral tensor-train decomposition
• Tensor Decompositions and Applications
• Tensor-train decomposition
• The alternating linear scheme for tensor optimization in the tensor train format
• The density-matrix renormalization group
• The density-matrix renormalization group in the age of matrix product states

Cited in

(5)

• Optimization landscape of Tucker decomposition
• Learning polynomial transformations via generalized tensor decompositions
• One-dimensional tensor network recovery
• Quaternion tensor left ring decomposition and application for color image inpainting
• Hankel tensor completion via two-layer optimization for functional magnetic resonance images recovery

This page was built for publication: Tensor ring decomposition: optimization landscape and one-loop convergence of alternating least squares

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5146616)