No label defined (Q4807667)

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 1914189
Language Label Description Also known as
default for all languages
No label defined
    English
    No label defined
    scientific article; zbMATH DE number 1914189

      Statements

      21 May 2003
      0 references
      Lanczos method
      0 references
      matrix eigenvalue problem
      0 references
      fast algorithm
      0 references
      minimal polynomial
      0 references
      stopping condition
      0 references
      round off error
      0 references
      three term recurrence relation
      0 references
      industrial application
      0 references
      finite difference
      0 references
      finite element
      0 references
      domain decomposition
      0 references
      0 references
      The book is self-contained, concise and describes the different developments of the method in the first part and its important applications in the second part. Recall that to solve the matrix eigenvalue problem, \(Au=\lambda u\) by the Lanczos method, a sequence of vectors \(b_{k}\) defined by the recursion \ \(b_{m}=\left( A-\alpha _{m-1}\right) b_{m-1}-\beta _{m-2}b_{m-2}\) is constructed, where at each stage of the iteration the coefficients \(\alpha _{k}\) and \(\beta _{k}\) are chosen to minimize the norm of the newly constructed vector \(b_{n}\). Simple algebraic formulae define the coefficients \(\alpha _{k}\) and \(\beta _{k}\) explicitly, which yields a fast algorithm. When \(b_{m}=0,\) the process stops and \(m\) is declared to be the order of the minimal polynomial. The main drawback, from the numerical point of view, is the stopping condition which may not be verified. In other words due to round off error, \(b_{m}\) may be very small but not zero. Thus it is necessary to use exact arithmetic instead of the floating point arithmetic. The sequences \(\alpha _{k}\) and \(\beta _{k}\) also define the minimal polynomial through a three term recurrence relation, \(P_{0}(\lambda)=1,\) \(P_{1}(\lambda)=\lambda -\alpha _{0},\) \(P_{n}(\lambda)=\left( \lambda -\alpha _{n-1}\right) P_{n-1}\left( \lambda \right) -\beta _{n-2}P_{n-2}(\lambda).\) The second part covers difficult industrial applications obtained from discretizing partial differential equations, finite difference and finite element equations. For example in chapter 7, the performance of the method on different partitions is presented. Also how a good speed-up is achieved when a frequency domain decomposition is used. At the end of the book, a brief biography of Cornelius Lanczos is included. In a clear and accessible style to a wide audience, the author explains the evolution of the method, its ramifications and various industrial applications.
      0 references

      Identifiers

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