Fast inversion of Chebyshev-Vandermonde matrices

Numerische Mathematik (Impact Factor: 1.55). 08/1995; DOI: 10.1007/s002110050018
Source: CiteSeer

ABSTRACT This paper contains two fast algorithms for inversion of Chebyshev-- Vandermonde matrices of the first and second kind. They are based on special representations of the Bezoutians of Chebyshev polynomials of both kinds. The paper also contains the results of numerical experiments which show that the algorithms proposed here are not only much faster, but also more stable than other algorithms available. It is also efficient to use the above two algorithms for solving Chebyshev--Vandermode systems of equations with preprocessing.

  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Jacobi (tridiagonal) and unitary Hessenberg matrices are among central objects in several areas and applications, of which three important ones are mentioned next. First, they were used in the design and analysis of a vast number of numerically accurate algorithms. Sec-ondly, Jacobi and unitary Hessenberg matrices capture short recurrence relations of real orthogonal polynomials and of the Szegö polynomials, respectively. Finally, they widely appear in signal processing applications, e.g., being related to lattice–ladder filter struc-tures. This thesis concerns a far-reaching generalization of the above two classical matrix classes, namely with quasiseparable matrices. Correspondingly, this thesis consists of three parts that study the role of quasiseparable matrices (i) in the design and analysis of numerically accurate algorithms, (ii) in orthogonal polynomial theory and (iii) in signal processing and digital filter design. Since quasiseparable matrices include both unitary Hessenberg and tridiagonal matrices as special cases, our approach allows one to obtain true and typically non-trivial generalizations of several classical results and algorithms. Quasiseparable matrices also arise in many applications in linear system theory and con-trol, statistics, mechanics, and others. This justifies the fact why quasiseparable matrices have been among the hottest research topics in Numerical Linear Algebra in recent years.
  • [Show abstract] [Hide abstract]
    ABSTRACT: In the present paper we use the displacement structure approach to introduce a new class of what are called confluent polynomial Vandermonde-like matrices, which generalize the ordinary simple nodes polynomial Vandermonde matrices studied earlier by different authors. The displacement structure approach leads to the explicit inversion formulas for confluent polynomial Vandermonde-like matrices and fast algorithms for inversion and for solving the associated linear systems.
    Linear Algebra and its Applications 05/2004; 382:61-82. DOI:10.1016/j.laa.2003.11.011 · 0.98 Impact Factor


Available from