Homotopic residual correction processes
The authors discuss a residual correction algorithm for the computation of the inverse or generalized inverse (in the sense of Moore-Penrose) of a matrix. The Newton iteration and some other residual correction processes for matrix inversion are initialized by means of the homotopic (continuation) method. The residual correction processes which are strongly numerically stable, allow effective parallel implementation and converge to the inverse, at least quadratically, if the initial approximation is close to the inverse. For a structured input matrix (e.g. Toeplitz, Hankel, Vandermonde, Cauchy or Pick matrices) the requirement of having a close initial approximation becomes very important. For Hermitian (or real symmetric) positive definite and indefinite matrices, the homotopic residual correction algorithm proposed here guarantees rapid convergence to the inverses, even where a not good initial approximation is available. Some additional practical convergence acceleration can be achieved by combining the homotopic algorithm with a linearly convergent iteration for the linear system. The homotopic residual correction processes do not require any pre-estimation of the smallest singular value of the input matrix. The homotopic algorithm is extended to the numerical computation of the Moore-Penrose generalized inverse and of the inverses of other type of matrices. Numerical experiments performed with general and structured matrices confirm the theory and provide the appropriate values of parameters which define the homotopic residual correction processes. The theoretical estimates for the number of homotopic residual correction steps grow in proportion to the logarithm of the condition number of the input matrix.
- A homotopic residual residual correction process
- scientific article; zbMATH DE number 3938435
- New homotopic/factorization and symmetrization techniques for Newton's and Newton/structured iteration
- scientific article; zbMATH DE number 270059
- An Improved Newton Iteration for the Generalized Inverse of a Matrix, with Applications
- A homotopic residual residual correction process
- A Note on an Iterative Method for Generalized Inversion of Matrices
- An Improved Newton Iteration for the Generalized Inverse of a Matrix, with Applications
- Approximate displacement rank and applications
- Computational methods of linear algebra
- Decreasing the Displacement Rank of a Matrix
- Displacement ranks of matrices and linear equations
- Fast Parallel Algorithms for QR and Triangular Factorization
- scientific article; zbMATH DE number 4213315 (Why is no real title available?)
- scientific article; zbMATH DE number 3671573 (Why is no real title available?)
- scientific article; zbMATH DE number 1350351 (Why is no real title available?)
- Inversion of Displacement Operators
- Iterative inversion of structured matrices
- Modular arithmetic for linear algebra computations in the real field
- Moore–Penrose Inversion of Square Toeplitz Matrices
- Newton's iteration for inversion of Cauchy-like and other structured matrices
- On Computations with Dense Structured Matrices
- On the Bezoutian Structure of the Moore–Penrose Inverses of Hankel Matrices
- On the Numerical Properties of an Iterative Method for Computing the Moore–Penrose Generalized Inverse
- Parallel solution of Toeplitzlike linear systems
- Stability of Methods for Solving Toeplitz Systems of Equations
- Structured matrices and Newton's iteration: unified approach
- Structured matrices and polynomials. Unified superfast algorithms
- New homotopic/factorization and symmetrization techniques for Newton's and Newton/structured iteration
- Structured matrices and Newton's iteration: unified approach
- An efficient computation of generalized inverse of a matrix
- Factorizations of hyperpower family of iterative methods via least squares approach
- Schur aggregation for linear systems and determinants
- A homotopic residual residual correction process
- New studies of randomized augmentation and additive preprocessing
- A continuation method for solving symmetric Toeplitz systems
- Chebyshev-type methods and preconditioning techniques
This page was built for publication: Homotopic residual correction processes
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5713233)
