Abstract
We present a method for implementing the ellipsoid algorithm, whose basic iterative step is a linear row manipulation on the matrix of inequalities. This step is somewhat similar to a simplex iteration, and may give a clue to the relation between the two algorithms. Geometrically, the step amounts to performing affine transformations which map the ellipsoids onto a fixed sphere. The method was tried successfully on linear programs with up to 50 variables, some of which required more than 24 000 iterations. Geometrical properties of the iteration suggest that the ellipsoid algorithm is numerically robust, which is supported by our computational experience.
Similar content being viewed by others
References
R.G. Bland, D. Goldfarb and M.J. Todd, “The ellipsoid method: A survey”,Operations Research 29 (1981) 1039–1091.
P. Gács and L. Lovász, “Khachiyan's algorithm for linear programming”,Mathematical Programming Study 14 (1981) 61–68.
D.B. Iudin and A.S. Nemirovskii, “Informational complexity and effective methods for the solution of convex extrema problems”,Ekonomika i Matematicheskie Metody 12 (1976) 351–369.
L.G. Khachiyan, “The polynomial algorithm in linear programming”,Zhurnal Vyschislitelnoi Matematiki i Matematicheskoi Fiziki 20 (1980) 51–68. Translated inUSSR Computational Mathematics and Mathematical Physics 20 (1980) 53–72.
B. Korte and R. Schrader, “A note on covergence proofs for Shor-Khachiyan methods”, Report 80156-OR. Institut für Ökonometrie und Operations Research, University of Bonn (Bonn 1980).
N.Z. Shor, “Convergence rate of the gradient descent method with dilation of the space”.Kibernetika 2 (1970) 80–85.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Halfin, S. The sphere method and the robustness of the ellipsoid algorithm. Mathematical Programming 26, 109–116 (1983). https://doi.org/10.1007/BF02591897
Received:
Accepted:
Issue Date:
DOI: https://doi.org/10.1007/BF02591897