The Strong Stability of Algorithms for Solving Symmetric Linear Systems (Q4206282)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 4126565
Language Label Description Also known as
default for all languages
No label defined
    English
    The Strong Stability of Algorithms for Solving Symmetric Linear Systems
    scientific article; zbMATH DE number 4126565

      Statements

      The Strong Stability of Algorithms for Solving Symmetric Linear Systems (English)
      0 references
      0 references
      0 references
      0 references
      1989
      0 references
      strong stability
      0 references
      error analysis
      0 references
      symmetric perturbations
      0 references
      symmetric matrices
      0 references
      Gaussian elimination with pivoting
      0 references
      symmetric systems
      0 references
      positive definite systems
      0 references
      An algorithm for solving linear equations of the form \(Ax=b\) is ``strongly stable'' on a class C of matrices if for every \(A\in C\) and for arbitrary vector b the computed solution \(\hat x\) solves a system \(\hat A\hat x=\hat b\) with \(\hat A\in C,\) which is near the original one (in the sense that \(\| \hat A-A\| /\| A\|\) and \(\| \hat b- b\| /\| b\|\) are ``small''). If we do not require the condition \(\hat A\in C,\) then we simply say that the algorithm is ``stable''. The concept of strong stability is important for instance in the error analysis in view of a new perturbation theorem about symmetric perturbations of symmetric matrices. In previous works the strong stability of certain algorithms for suitable classes of matrices was established. Here a general theorem is proved on this subject. It states that if a method is stable for the class of nonsingular symmetric matrices or the class of symmetric positive definite matrices, then it is strongly stable for the same class. As new application of this result it is shown that Gaussian elimination with pivoting and \textit{J. O. Aasen}'s method [BIT 11, 233-242 (1971; Zbl 0242.65032)] are strongly stable when applied to symmetric systems. Other applications in the class of positive definite systems are briefly discussed.
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references