A Matrix Factorization and Its Application to Large-Scale Linear Programming
Abstract
As an alternative to the LU matrix factorization, we consider a factorization that uses the lower triangular part of the original matrix as one factor and computes the other factors as a product of rank-one update matrices. After reviewing and improving the time complexity, the requirements, the stability and the efficiency of this method, we derive a stable factorization algorithm which we implement in FORTRAN77 within the framework of the simplex algorithm for linear programming. A comparison of our numerical results with those obtained through the code MINOS 5.3 indicate that our method may be more efficient than an ordinary LU decomposition for some matrices whose order ranges between 28 and 1481, especially when these matrices are almost triangular.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1989
- Accession Number
- ADA211293
Entities
People
- Pierre F. Demazancourt
Organizations
- Stanford University