Scribd
Upload
22 views11 pages

Matrix Iteration Methods Overview

Matrix iteration methods are iterative algorithms for solving linear systems and computing eigenvalues/eigenvectors. They are categorized into classical, Krylov, block, conjugate gradient, dense, and projection methods, with specialized and hybrid techniques for large-scale problems. The choice of method is influenced by the problem's size, sparsity, and accuracy needs.

Uploaded by

Rojan Mathew
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
22 views11 pages

Matrix Iteration Methods Overview

Matrix iteration methods are iterative algorithms for solving linear systems and computing eigenvalues/eigenvectors. They are categorized into classical, Krylov, block, conjugate gradient, dense, and projection methods, with specialized and hybrid techniques for large-scale problems. The choice of method is influenced by the problem's size, sparsity, and accuracy needs.

Uploaded by

Rojan Mathew
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd

Classification of Matrix Iteration

Methods
Detailed Categorization
Introduction
• Matrix iteration methods are iterative
numerical algorithms designed for solving
linear systems, computing
eigenvalues/eigenvectors, and handling large-
scale sparse matrices.

• This presentation provides a structured


classification of these methods.
Classical Power-Based Methods
• • Standard Power Iteration:
• - Iteration: x_{k+1} = A x_k / ||A x_k||
• - Finds dominant eigenvalue.

• • Inverse Iteration:
• - Iteration: (A - μI) y = x_k, then x_{k+1} = y /
||y||
• - Finds eigenvalue near shift μ.
Krylov Subspace Methods
• • Lanczos Algorithm:
• - Builds tridiagonal matrix T from Krylov
basis.
• - Efficient for symmetric matrices.

• • Arnoldi Iteration:
• - Generalization of Lanczos for non-
symmetric matrices.
• - Produces Hessenberg form.
Block and Subspace Methods
• • Subspace Iteration:
• - x_{k+1} = A X_k, then orthonormalize.
• - Finds multiple eigenvalues simultaneously.

• • Simultaneous Iteration:
• - Parallel version of subspace iteration.

• • LOBPCG:
• - Uses block conjugate gradient with
Conjugate Gradient-Based
Methods
• • Conjugate Gradient for Eigenvalues:
• - Minimizes Rayleigh quotient in CG
direction.

• • Generalized CG Methods:
• - Extends to non-symmetric or indefinite
problems.
Dense Matrix Methods
• • QR Algorithm:
• - Repeated QR decomposition: A_{k+1} = R_k
Q_k.
• - Converges to Schur form.

• • Jacobi Method:
• - Iterative rotation of matrix entries.
• - Diagonalizes symmetric matrices.
Projection and Davidson-Type
Methods
• • Davidson Method:
• - Iterative correction using diagonal
preconditioning.

• • Jacobi-Davidson:
• - Solves correction equation with
orthogonalization.

• • Chebyshev-Davidson Hybrid:
Specialized Krylov Methods for
Eigenvalue Problems
• • GMRES-Based Methods:
• - Adapts GMRES for eigenvalue
approximations.

• • BiCGSTAB Variants:
• - Handles non-symmetric eigenvalue
problems efficiently.
Advanced and Hybrid Techniques
• • Deflation Techniques:
• - Removes converged eigenvalues to
compute next.

• • Polynomial Filtering:
• - Uses polynomial filters to enhance
convergence of specific eigenvalues.

• • Preconditioned Methods:
Summary
• • Matrix iteration methods span classical,
Krylov, block, conjugate gradient, dense, and
projection approaches.
• • They include specialized and hybrid
techniques for large-scale computations.
• • Method selection depends on problem size,
sparsity, and accuracy requirements.

You might also like