A Geometric Newton Method for Oja's Vector Field

DOI10.1162/NECO.2008.04-08-749MaRDI QIDQ3628017zbMATH OpenOpenAlexWikidataFDO

Authors P.-A. Absil, Mariya Ishteva, Lieven De Lathauwer, Sabine Van Huffel

Publication date 19 May 2009

Published in Neural Computation (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/0804.0989

zbMATH Keywords

numerical experimentsmatrix equationssuperlinear convergencegeometric Newton methodOja's flowdegenerate zeros

Mathematics Subject Classification ID

Numerical computation of solutions to systems of equations (65H10) Matrix equations and identities (15A24)

Abstract: Newton's method for solving the matrix equation $F(X)equivAX-X{X}^{T}AX=0$ runs up against the fact that its zeros are not isolated. This is due to a symmetry of $F$ by the action of the orthogonal group. We show how differential-geometric techniques can be exploited to remove this symmetry and obtain a ``geometric Newton algorithm that finds the zeros of $F$. The geometric Newton method does not suffer from the degeneracy issue that stands in the way of the original Newton method.

Recommendations

• Newton's methods from a geometric point of view
• Newton's method for approximating zeros of vector fields on Riemannian manifolds
• Newton's method, zeroes of vector fields, and the Riemannian center of mass
• A unifying local convergence result for Newton's method in Riemannian manifolds
• Optimization algorithms on the Grassmann manifold with application to matrix eigenvalue problems

Cites work

• scientific article; zbMATH DE number 3138903 (Why is no real title available?)
• scientific article; zbMATH DE number 3928227 (Why is no real title available?)
• scientific article; zbMATH DE number 3381785 (Why is no real title available?)
• scientific article; zbMATH DE number 961607 (Why is no real title available?)
• scientific article; zbMATH DE number 3194988 (Why is no real title available?)
• A dual purpose principal and minor component flow
• A simplified neuron model as a principal component analyzer
• Adaptive eigenvalue computations using Newton's method on the Grassmann manifold
• Analysis of a class of continuous-time algorithms for principal component analysis and subspace tracking
• Choosing the Forcing Terms in an Inexact Newton Method
• Cubically Convergent Iterations for Invariant Subspace Computation
• Newton's method on Riemannian manifolds and a geometric model for the human spine
• On the largest principal angle between random subspaces
• Pseudo-Transient Continuation for Nonsmooth Nonlinear Equations
• Riemannian geometry of Grassmann manifolds with a view on algorithmic computation
• Stability and Convergence of Principal Component Learning Algorithms
• The Geometry of Algorithms with Orthogonality Constraints
• Trust Region Algorithms and Timestep Selection

Cited in

(8)

• Solving PhaseLift by Low-Rank Riemannian Optimization Methods for Complex Semidefinite Constraints
• A generalized geometric spectral conjugate gradient algorithm for finding zero of a monotone tangent vector field on a constant curvature Hadamard manifold
• Quotient geometry with simple geodesics for the manifold of fixed-rank positive-semidefinite matrices
• Restricted Riemannian geometry for positive semidefinite matrices
• A Riemannian nonmonotone spectral method for self-adjoint tangent vector field
• Computing laser beam paths in optical cavities: an approach based on geometric Newton method
• A Riemannian derivative-free Polak-Ribiére-Polyak method for tangent vector field
• Grassmann algorithms for low rank approximation of matrices with missing values

This page was built for publication: A Geometric Newton Method for Oja's Vector Field

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3628017)