Treffer: On the normal matrix of the polynomial LS problem over the Chebyshev points
Title:
On the normal matrix of the polynomial LS problem over the Chebyshev points
Authors:
Source:
Linear Algebra and its Applications. 378:61-69
Publisher Information:
Elsevier BV, 2004.
Publication Year:
2004
Subject Terms:
Numerical Analysis, Numerical solutions to overdetermined systems, pseudoinverses, Algebra and Number Theory, Polynomial approximation, least squares method, Vandermonde matrices, Direct numerical methods for linear systems and matrix inversion, 01 natural sciences, Factorization of matrices, Approximation by polynomials, Computation of special functions and constants, construction of tables, Vandermonde matrix, polynomial approximation, Discrete Mathematics and Combinatorics, Geometry and Topology, 0101 mathematics, numerical experiments, Real polynomials: analytic properties, etc, Conditioning, Cholesky factorization
Document Type:
Fachzeitschrift
Article
File Description:
application/xml
Language:
English
ISSN:
0024-3795
DOI:
10.1016/j.laa.2003.09.002
Access URL:
https://zbmath.org/2081746
https://doi.org/10.1016/j.laa.2003.09.002
https://iris.unical.it/handle/20.500.11770/158941
https://www.sciencedirect.com/science/article/pii/S0024379503007493
https://www.sciencedirect.com/science/article/abs/pii/S0024379503007493
https://core.ac.uk/display/82438528
https://hdl.handle.net/20.500.11770/158941
https://doi.org/10.1016/j.laa.2003.09.002
https://iris.unical.it/handle/20.500.11770/158941
https://www.sciencedirect.com/science/article/pii/S0024379503007493
https://www.sciencedirect.com/science/article/abs/pii/S0024379503007493
https://core.ac.uk/display/82438528
https://hdl.handle.net/20.500.11770/158941
Rights:
Elsevier Non-Commercial
Accession Number:
edsair.doi.dedup.....a261cabe7675fe1d43e18e66cfd77b2b
Database:
OpenAIRE
Weitere Informationen
Let \(V = [x_i^{j-1}]\) be an \(n\times m\) (\(n\geq m\)) Vandermonde matrix with the Chebyshev nodes \(x_i = \cos(\pi (2i-1) /2n)\). This paper provides an explicit formula for the Cholesky factor of the matrix \(V^T V\), leading to an \(O(m (m+n) )\) method for solving polynomial approximation problems. It turns out that the entries of the Cholesky factor and its inverse can be represented as integers divided by powers of two. Numerical experiments suggest that this approach is more efficient but also sometimes less accurate than methods based on orthogonal polynomials for solving approximation problems.