A Strictly Improving Phase 1 Algorithm Using Least-Squares Subproblems
Abstract
Although the simplex method's performance in solving linear programming problems is usually quite good, it does not guarantee strict improvement at each iteration on degenerate problems. Instead of trying to recognize and avoid degenerate steps in the simplex method (as some variants do) , we have developed a new Phase I algorithm that is completely impervious to degeneracy, with strict improvement attained at each iteration. it is also noted that the new Phase I algorithm is closely related to a number of existing algorithms. When tested on the 30 smallest NETLIB linear programming test problems, the computational results for the new Phase I algorithm were almost 3. 5 times faster than the simplex method; on some problems, it was over 10 times faster.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1992
- Accession Number
- ADA251913
Entities
People
- G. B. Dantzig
- J. W. Davis
- S. A. Leichner
Organizations
- Stanford University