A geometric perspective on the singular value decomposition

DOI10.13137/0049-4704/11222MaRDI QIDQ2809909zbMATH OpenOpenAlexFDO

Authors Raffaella Paoletti, Giorgio Ottaviani

Publication date 30 May 2016

Published in Rendiconti dell'Istituto di Matematica dell'Università di Trieste (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1503.07054

zbMATH Keywords

best rank \(r\) approximationcritical pointsEuclidean distance degreesecant varietiessingular space of a tensorsingular value decompositiontangent spacetensors

Mathematics Subject Classification ID

Projective techniques in algebraic geometry (14N05) Eigenvalues, singular values, and eigenvectors (15A18) Multilinear algebra, tensor calculus (15A69) Numerical solutions to overdetermined systems, pseudoinverses (65F20)

Abstract: This is an introductory survey, from a geometric perspective, on the Singular Value Decomposition (SVD) for real matrices, focusing on the role of the Terracini Lemma. We extend this point of view to tensors, we define the singular space of a tensor as the space spanned by singular vector tuples and we study some of its basic properties.

Review text

The authors start from the characterization of the singular value decomposition (SVD) of a matrix as being given by the best rank \(1\) approximation and from the characterization of the best rank \(r\) approximation. Then, they introduce geometric concepts like secant varieties, their tangent spaces, and critical points of the distant function of a matrix and varieties. This allows for a redefinition and a new proof of the SVD leading also to its coordinate-free version. Continuing, they consider singular vector tuples and singular spaces of tensors and prove a result on (symmetric) tensor decompositions for general (symmetric) tensors based here on earlier work of \textit{S. Friedland} and \textit{G. Ottaviani} [Found. Comput. Math. 14, No. 6, 1209--1242 (2014; Zbl 1326.15036)]. Finally, they give a definition of the Euclidean distance degree and higher-order SVD and formulate generalizations of their former results using these concepts.

Recommendations

• A Geometrically Inspired Proof of the Singular Value Decomposition
• Making matrices better: geometry and topology of singular value and polar decomposition
• A Multilinear Singular Value Decomposition
• scientific article; zbMATH DE number 782060
• Rank-one approximation to high order tensors

Cited in

(18)

• Skew symmetric logarithms and geodesics on \(\mathrm{O}_n(\mathbb{R})\)
• The Euclidean distance degree of orthogonally invariant matrix varieties
• Some remarks on the Jordan-Chevalley decomposition
• Best rank \(k\) approximation for binary forms
• Asymptotics of degrees and ED degrees of Segre products
• Tensors with eigenvectors in a given subspace
• The critical space for orthogonally invariant varieties
• Best rank-k approximations for tensors: generalizing Eckart-Young
• Geometry-aware principal component analysis for symmetric positive definite matrices
• Geometry of curves in \(\mathbb{R}^n\) from the local singular value decomposition
• Exact solutions in low-rank approximation with zeros
• Quartic coincidences and the singular value decomposition
• On Tensors That Are Determined by Their Singular Tuples
• Making matrices better: geometry and topology of singular value and polar decomposition
• Hypocoercivity and controllability in linear semi‐dissipative Hamiltonian ordinary differential equations and differential‐algebraic equations
• The span of singular tuples of a tensor beyond the boundary format
• Generalized principal logarithms and Riemannian properties of a class of subgroups of \(\mathbf{U_n}\) endowed with the Frobenius bi-invariant metric
• Euclidean distance degree in manifold optimization

Describes a project that uses

Uses Software

• Macaulay2

This page was built for publication: A geometric perspective on the singular value decomposition

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2809909)