The Convergence of the Method of Conjugate Gradients at Isolated Extreme Points of the Spectrum,

Abstract

Let A be a positive definite matrix with a simple eigenvalue (lambda sub 1) that lies outside an interval (alpha,beta) containing the remaining eigenvalues. Let the method of conjugate gradients be applied to the solution of the linear system Az = b producing a sequence of iterates z sub 0,z sub 1,... and an associated sequence of error vectors (e sub i) = z - (z sub i). In the paper bounds are obtained for the component of the error vector lying along the eigenvector associated with (lambda sub 1). The bounds imply that, provided (lambda sub 1) is well separated from (alpha,beta), this component will decrease rapidly even when the matrix A is moderately ill conditioned. (Author)

Document Details

Document Type
Technical Report
Publication Date
Aug 01, 1973
Accession Number
AD0765748

Entities

People

  • G. W. Stewart

Organizations

  • Carnegie Mellon University

Tags

DTIC Thesaurus Topics

  • Convergence
  • Eigenvalues
  • Eigenvectors
  • Intervals
  • Linear Systems
  • Mathematics
  • Sequences
  • Spectra

Readers

  • Analytical Mechanics
  • Approximation Theory.
  • Regression Analysis.