Advertisement

Numerical Algorithms

, Volume 8, Issue 2, pp 241–268 | Cite as

Estimates in quadratic formulas

  • Gene H. Golub
  • Zdeněk Strakoš
Article

Abstract

LetA be a real symmetric positive definite matrix. We consider three particular questions, namely estimates for the error in linear systemsAx=b, minimizing quadratic functional minx(x T Ax−2b T x) subject to the constraint ‖x‖=α, α<‖A−1b‖, and estimates for the entries of the matrix inverseA−1. All of these questions can be formulated as a problem of finding an estimate or an upper and lower bound onu T F(A)u, whereF(A)=A−1 resp.F(A)=A−2,u is a real vector. This problem can be considered in terms of estimates in the Gauss-type quadrature formulas which can be effectively computed exploiting the underlying Lanczos process. Using this approach, we first recall the exact arithmetic solution of the questions formulated above and then analyze the effect of rounding errors in the quadrature calculations. It is proved that the basic relation between the accuracy of Gauss quadrature forf(λ)=λ−1 and the rate of convergence of the corresponding conjugate gradient process holds true even for finite precision computation. This allows us to explain experimental results observed in quadrature calculations and in physical chemistry and solid state physics computations which are based on continued fraction recurrences.

Keywords

Solid State Conjugate Gradient Continue Fraction Quadratic Formula Real Vector 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. [1]
    K. Amos, L. Berge, F.A. Brieva, A. Katsogiannis, L. Petris and L. Rikus, Off-shell properties of the two-nucleonT matrix, Phys. Rev. C37 (1988) 934–948.Google Scholar
  2. [2]
    P. Deufhard, Cascadic conjugate gradient methods for elliptic partial differential equations I. Algorithm and numerical results, preprint SC93-23, Konrad Zuse Zentrum, Berlin (1993).Google Scholar
  3. [3]
    G. Dahlquist, S.C. Eisenstat and G.H. Golub, Bounds for the error of linear systems of equations using the theory of moments, J. Math. Anal. Appl. 37 (1972) 151–166.Google Scholar
  4. [4]
    G. Dahlquist, G.H. Golub and S.G. Nash, Bounds for the error in linear systems, in:Proc. Workshop on Semi-Infinite Programming, ed. R. Hettich (Springer, 1978) pp. 154–172.Google Scholar
  5. [5]
    P.J. Davis and P. Rabinowitz,Numerical Integration (Academic Press, 1984).Google Scholar
  6. [6]
    B. Fischer and G.H. Golub, On the error computation for polynomial based iterative methods, Manuscript NA 92-21, Dept. of Comp. Sc., Stanford University (1992).Google Scholar
  7. [7]
    B. Fischer and R. Freund, On adaptive weighted polynomial preconditioning for Hermitian positive definite matrices, SIAM J. Sci. Comp. (1993) to appear.Google Scholar
  8. [8]
    R. Freund and M. Hochbruck, Gauss quadratures associated with the Arnoldi process and the Lanczos algorithm, In:Linear Algebra for Large Scale and Real-Time Applications, eds. M.S. Moonen et al. (Kluwer Academic, Amsterdam, 1993) pp. 377–380.Google Scholar
  9. [9]
    W. Gautschi, Construction of Gauss-Christoffel quadrature formulas, Math. Comp. 22 (1968) 251–270.Google Scholar
  10. [10]
    W. Gautschi, A survey of Gauss-Christoffel quadrature formulae, in:E.B. Christoffel — The Influence of His Work on Mathematics and the Physical Sciences, eds. P.L. Bultzer and F. Fehér (Birkhäuser, Boston, 1981) pp. 73–157.Google Scholar
  11. [11]
    W. Gautschi, On generating orthogonal polynomials, SIAM J. Sci. Stat. Comp. 3 (1982) 289–317.Google Scholar
  12. [12]
    G.H. Golub and U.v. Matt, Quadratically constrained least squares and quadratic problems, Numer. Math. 59 (1991) 561–580.Google Scholar
  13. [13]
    G.H. Golub and G. Meurant, Matrices, moments and quadrature, Manuscript SCCM93-07, Dept. of Comp. Sc., Stanford University (1993).Google Scholar
  14. [14]
    G.H. Golub and J.H. Welsch, Calculation of Gauss quadrature rules, Math. Comp. 23 (1969) 221–230.Google Scholar
  15. [15]
    G.H. Golub, Some modified matrix eigenvalue problems, SIAM Rev. 15 (1973) 318–334.Google Scholar
  16. [16]
    G.H. Golub, Bounds for matrix moments, Rocky Mountain J. Math. 4 (1974) 207–211.Google Scholar
  17. [17]
    G.H. Golub and M.D. Kent, Estimates of eigenvalues for iterative methods, Math. Comp. 53 (1989) 619–626.Google Scholar
  18. [18]
    A. Greenbaum, Behavior of slightly perturbed Lanczos and Conjugate Gradient recurrences, Lin. Alg. Appl. 113 (1989) 7–63.Google Scholar
  19. [19]
    A. Greenbaum and Z. Strakoš, Predicting the behavior of finite precision Lanczos and Conjugate Gradient computations, SIAM J. Matrix Anal. Appl. 13 (1992) 121–137.Google Scholar
  20. [20]
    G.H. Golub and Z. Strakoš, Estimates in quadratic formulas, Manuscript SCCM93-08, Dept. of Comp. Sci., Stanford University (1993).Google Scholar
  21. [21]
    L.A. Hageman and D.M. Young,Applied Iterative Methods (Academic Press, 1981).Google Scholar
  22. [22]
    M.R. Hestenes and E. Stiefel, Methods of Conjugate Gradients for solving linear systems. J. Res. Natl. Bureau of Standards 49 (1952) 409–436.Google Scholar
  23. [23]
    J. Horáček and T. Sasakawa, Method of continued fractions with application to atomic physics, Phys. Rev. A30 (1984) 2274–2277.Google Scholar
  24. [24]
    J. Horáěk and T. Sasakawa, Method of continued fractions for on- and off-shellT matrix of local and nonlocal potentials, Phys. Rev. C37 (1985) 70–75.Google Scholar
  25. [25]
    J. Horáček, A note on iterative solution of scattering equations, Acta Phys. Polonica A71 (1987) 3–8.Google Scholar
  26. [26]
    J. Horáček, Extrapolation technique for solution of scattering integral equations, J. Phys. A: Math. Gen. 22 (1989) 1–9.Google Scholar
  27. [27]
    L. Knizhnerman, The simple Lanczos process: error estimates for the Gaussian quadrature formula and their applications, (1994), in preparation.Google Scholar
  28. [28]
    M.D. Kent, Chebyshev, Krylov and Lanczos, matrix relationship and computations, Ph.D. Thesis, Stanford University, Stanford (1989).Google Scholar
  29. [29]
    J. Kautský and G.H. Golub, On the calculation of Jacobi matrices, Lin. Alg. Appl. 52/53 (1983) 439–455.Google Scholar
  30. [30]
    S. Karlin and L.S. Shapley,Geometry of Moment Spaces, AMS Memoirs 12 (AMS, Providence, 1953)Google Scholar
  31. [31]
    G. Moro and J.H. Freed, The Lanczos algorithm in molecular dynamics: calculation of spectral densities, in:Large Scale Eigenvalue Problems, eds. J. Cullum and R.A. Willoughby (Elsevier/North Holland, Amsterdam, 1986) pp. 143–161.Google Scholar
  32. [32]
    H.D. Mayer and S. Pal, A band-Lanczos method for computing matrix elements of a resolvent, J. Chem. Phys. 91 (1989) 6195–6204.Google Scholar
  33. [33]
    Y. Notay, On the convergence rate of the Conjugate Gradients in presence of rounding errors, Numer. Math. (1993), to appear.Google Scholar
  34. [34]
    B.N. Parlett, Misconvergence in the Lanczos algorithm, Res. Rep. PAM-404, U. of California, Berkeley (1987).Google Scholar
  35. [35]
    C.C. Paige, The computation of eigenvalues and eigenvectors of very large sparse matrices, Ph.D. Thesis, U. of London (1971).Google Scholar
  36. [36]
    C.C. Paige, Computational variants of the Lanczos method for the eigenproblem, J. Inst. Math. Appl. 10 (1972) 373–381.Google Scholar
  37. [37]
    C.C. Paige, Error analysis of the Lanczos algorithm for tridiagonalizing a symmetric matrix, J. Inst. Math. Appl. 18 (1976) 341–349.Google Scholar
  38. [38]
    C.C. Paige, Accuracy and effectiveness of the Lanczos algorithm for the symmetric eigenproblem, Lin. Alg. Appl. 34 (1980) 235–258.Google Scholar
  39. [39]
    W. Reinhardt,L 2 discretization of atomic and molecular electronic continua: moment, quadrature andJ-matrix techniques, Comp. Phys. Commun. 17 (1979) 1–21.Google Scholar
  40. [40]
    P.D. Robinson and A.J. Wathen, Variational bounds on the entries of the inverse of a matrix, IMA J. Numer. Anal. 12 (1992).Google Scholar
  41. [41]
    E. Stiefel, Relaxationsmethoden bester Strategie zur Lösung Linearer Gleichungsysteme, Comm. Math. Helv. 29 (1955) 157–179.Google Scholar
  42. [42]
    H.D. Simon, The Lanczos algorithm with partial reorthogonalization, Math. Comp. 42 (1984) 115–136.Google Scholar
  43. [43]
    H.D. Simon, Analysis of the symmetric Lanczos algorithm with reorthogonalization methods, Lin. Alg. Appl. 61 (1984) 101–131.Google Scholar
  44. [44]
    Z. Strakoš, On the real convergence rate of the Conjugate Gradient method, Lin. Alg. Appl. 154–156 (1991) 535–549.Google Scholar
  45. [45]
    Z. Strakoš and A. Greenbaum, Open questions in the convergence analysis of the Lanczos process for the real symmetric eigenvalue problem, IMA Preprint Series 934, IMA, U. of Minnesota (1992).Google Scholar
  46. [46]
    G. Szegö,Orthogonal Polynomials, AMS Colloq. Publ. 23 (AMS, Providence, 1939).Google Scholar
  47. [47]
    H.S. Wilf,Mathematics for the Physical Sciences (Wiley, 1962).Google Scholar

Copyright information

© J.C. Baltzer AG, Science Publishers 1994

Authors and Affiliations

  • Gene H. Golub
    • 1
  • Zdeněk Strakoš
    • 2
  1. 1.Department of Computer ScienceStanford UniversityStanfordUSA
  2. 2.Institute of Computer ScienceAcademy of Science of the Czech RepublicPraha 8Czech Republic

Personalised recommendations