📓 CME 302

        • Drawing 2023-10-04 2.excalidraw
        • Drawing 2023-10-04 12.13.49.excalidraw
        • Drawing 2023-10-04 12.26.03.excalidraw
        • Drawing 2023-10-15 18.35.05.excalidraw
        • Drawing 2023-10-15 18.36.19.excalidraw
        • Drawing 2023-10-18 11.37.23.excalidraw
        • GMRES least-squares problem 2023-11-29 10.49.27.excalidraw
        • Gram-Schmidt 2023-10-15 17.54.00.excalidraw
        • Least-squares solution using SVD 2023-10-15 20.28.15.excalidraw
        • MINRES 2023-12-04 08.40.09.excalidraw
        • QR iteration with shift 2023-10-30 11.49.00.excalidraw
        • QR using Givens transformations 2023-10-18 11.41.30.excalidraw
        • QR using Householder transformations 2023-10-18 11.37.36.excalidraw
      • 2024 complete and painless Conjugate Gradient
      • Accelerating convergence using a shift
      • Algorithm for QR iteration with shift
      • Algorithm for the Arnoldi process
      • Algorithm for the Lanczos process
      • All the orthogonality relations in CG
      • Angle between subspaces
      • Applying a Householder transformation
      • Arnoldi process
      • Backward error analysis for LU
      • Bootcamp
      • Brief introduction to Conjugate Gradients
      • Cauchy-Schwarz
      • CG search directions
      • Chebyshev iteration
      • Cholesky factorization
      • Classical iterative methods to solve sparse linear systems
      • Computational cost of Arnoldi and Lanczos
      • Computationally efficient search directions
      • Computing eigenvalues
      • Computing eigenvectors using the Schur decomposition
      • Computing multiple eigenvalues
      • Conditioning of a linear system
      • Conjugate Gradients algorithm
      • Conjugate Gradients code
      • Conjugate Gradients Version 1
      • Connection between Arnoldi and polynomials of A
      • Connection between Krylov subspace and Arnoldi
      • Convergence of classical iterative methods
      • Convergence of GMRES
      • Convergence of Lanczos eigenvalues for symmetric matrices
      • Convergence of Lanczos inner eigenvalues
      • Convergence of the Conjugate Gradients
      • Convergence of the orthogonal iteration
      • Convergence with conjugate steps
      • Deflation in the QR iteration
      • Determinant
      • Diagonalizable matrices
      • Dot product
      • Eigenvalues
      • Eigenvalues cannot be computed exactly
      • Exact shift
      • Existence of LU
      • Existence of the Cholesky factorization
      • Flexible preconditioned conjugate gradient method
      • Floating point arithmetic and unit roundoff error
      • Floating point arithmetic is different from regular arithmetic
      • Floating point numbers
      • Forward and backward error
      • Gauss-Seidel iteration
      • Ghost eigenvalues in the Lanczos process
      • GMRES
      • GMRES algorithm
      • GMRES least-squares problem
      • Gram-Schmidt
      • Hermitian and symmetric matrices
      • Householder transformation
      • Invertible matrix
      • Iterative methods for eigenvalue computation
      • Jacobi iteration
      • Key idea of iterative methods for eigenvalue computation
      • Krylov iterative methods to solve sparse linear systems
      • Krylov methods for sparse systems
      • Krylov subspace
      • Lanczos process
      • Least-squares problems
      • Least-squares solution using QR
      • Least-squares solution using SVD
      • LU algorithm
      • LU and determinant
      • Matrix block operations
      • Matrix-vector and matrix-matrix product
      • Method of normal equation
      • Method of power iteration
      • MINRES
      • Motivation of iterative methods for eigenvalue computation
      • Operator and matrix norms
      • Optimal step size
      • Orthogonal iteration
      • Orthogonal iteration algorithm
      • Orthogonal matrix and projector
      • Outer form of matrix-matrix product
      • Preconditioned Conjugate Gradients algorithm
      • Preconditioned Conjugate Gradients code
      • Preconditioning
      • Preconditioning the Conjugate Gradients algorithm
      • Projection
      • Pythagorean theorem
      • QR factorization
      • QR factorization and least-squares
      • QR iteration
      • QR iteration 2x2 example
      • QR iteration for upper Hessenberg matrices
      • QR iteration with shift
      • QR using Givens transformations
      • QR using Householder transformations
      • Rajat's painless Conjugate Gradients
      • Residuals and solution increments in CG
      • Row pivoting
      • Schur decomposition
      • Sensitivity analysis
      • Sherman-Morrison-Woodbury formula
      • Singular value decomposition
      • Solving linear systems
      • Solving linear systems using LU
      • Solving triangular systems
      • Some orthogonality relations in CG
      • SOR iteration
      • SOR iteration as a splitting method
      • Space and time costs of CG and GMRES
      • Splitting methods
      • Stability of the Cholesky factorization
      • Stability of the LU factorization
      • Subspace and linear independence
      • Summary of convergence and cost of the QR iteration
      • Summary of least-squares solution methods
      • Symmetric and unsymmetric QR iteration
      • Symmetric Positive Definite Matrices
      • The four fundamental spaces
      • Three-term recurrence
      • Trace
      • Triangular factorization
      • Uniqueness of the QR factorization
      • Unitarily diagonalizable matrices
      • Upper Hessenberg form for the QR iteration
      • Vector norms
      • Vectors and matrices
      • Why eigenvalues
    Home

    ❯

    Stability of the Cholesky factorization

    Stability of the Cholesky factorization

    Dec 05, 20241 min read

    The Cholesky factorization algorithm is very stable. The entries of L cannot grow. This is true even if we do not pivot.

    This can be seen from the equations:

    A=LLT,j∑​lij2​=aii​

    The entries of L are bounded by the square root of the diagonal entries of A. From our backward error analysis, this implies that the factorization is backward stable.

    Stability of the LU factorization, Backward error analysis for LU, Row pivoting, Cholesky factorization, Existence of the Cholesky factorization


    Graph View

    Backlinks

    • Method of normal equation
    • Solving linear systems

    Created with Quartz v4.3.1 © 2024

    • GitHub
    • Canvas