Probabilistic Analysis of Gaussian Elimination Without Pivoting
The authors present a probabilistic analysis of the numerical stability of the Gaussian elimination without pivoting. Using the normal distribution with zero mean and unit variance for the entries of \(A\), density functions for the entries of \(L\) and \(U\) (when \(A=LU)\) are derived. It is proved that the probability of the occurence of a pivot less than \(\varepsilon\) in magnitude is \(O(\varepsilon)\). Bounds on the probabilities of large growth factors are derived. In particular it is proved that the probabilities \(\text{Prob} ({|U|_\infty \over|A|_\infty} >m^{2.5})\) and \(\text{Prob} (|L|_\infty >m^3)\) decay algebraically to zero as \(m\) tends to infinity. Finally numerical experiments are presented to support the theoretical results.
- Probabilistic analysis of complex Gaussian elimination without pivoting
- The 70th anniversary of the distribution of random matrices: A survey
- Numerically safe Gaussian elimination with no pivoting
- How bad are Vandermonde matrices?
- scientific article; zbMATH DE number 5872172 (Why is no real title available?)
- Smoothed Analysis of the Condition Numbers and Growth Factors of Matrices
- Average-Case Stability of Gaussian Elimination
- scientific article; zbMATH DE number 3928192 (Why is no real title available?)
- Error accumulation in solving sparse systems of linear algebraic equations with a principal diagonal
- Average-case analysis of the Gaussian elimination with partial pivoting
- Distribution of the number of pivots needed using Gaussian elimination with partial pivoting on random matrices
- LU Factorization of Discrete Random Matrices
- On the Growth Factor of Random Matrices
This page was built for publication: Probabilistic Analysis of Gaussian Elimination Without Pivoting
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4337395)
