Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

A fast LU decomposition algorithm for the sparse revised simplex method

Conference ·
OSTI ID:54443
 [1];  [2]
  1. State Univ. of New York, Buffalo, NY (United States)
  2. Transtech Parallel Systems Corp., Ithaca, NY (United States)

Many algorithms reduce fill-ins in LU decomposition by rearranging columns and rows of the sparse matrix. We take a different approach for the LU decomposition in the revised simplex algorithm. The basis matrix in this algorithm is extremely sparse and there are many column and row singletons. We propose a new algorithm to rearrange the column and row singletons iteratively. This algorithm reduces the size of the submatrix to be factorized to minimal. It reduces fill-ins and computation time drastically. This algorithm has been tested on many realistic data from netlib, ranging from small-scale to large-scale.

OSTI ID:
54443
Report Number(s):
DOE/ER/25151--1-Vol.1; CONF-930331--Vol.1; CNN: Contract DABT63-91-k-0005
Country of Publication:
United States
Language:
English

Similar Records

Multiprocessor sparse LU decomposition with controlled fill-in
Thesis/Dissertation · Tue Dec 31 23:00:00 EST 1985 · OSTI ID:6891551

An approach for parallelizing any general unsymmetric sparse matrix algorithm
Conference · Thu Nov 30 23:00:00 EST 1995 · OSTI ID:125544

Computing row and column counts for sparse QR and LU factorization
Journal Article · Sun Dec 31 23:00:00 EST 2000 · BIT Numerical Mathematics · OSTI ID:834431