Decomposition of integer programs and of generating sets
In this paper we investigate techniques for decomposing the matrix of coefficients of a family of integer programs. From a more practical point of view, these techniques are useful to design a primal algorithm that solves the integer program via generating sets. In this con text our approach is applied to the Frobenius problem and to integer programming instances that seem to be difficult for LP-based integer programming codes. From a theoretical point of view, the techniques for decomposing a matrix that we present in this paper give rise to bounds on the L1-norm of all the elements in the Hilbert basis of a pointed cone. Moreover, applying our decomposition techniques we can show that any 0/1 linear integer program with a fixed number of constraints and a fixed number of digits to encode each coefficient in the matrix can be solved in polynomial time. A relation of our method to the group theoretic approach exists and is discussed as well.
Keywordsgenerating set group theoretic approach integer programming test set scaling primal method knapsack problem
Unable to display preview. Download preview PDF.
- 1.A. Brauer & J. E. Shockley: On a problem of Frobenius, Journal für reine und angewandte Mathematik 211 (1962) 399–408Google Scholar
- 2.P. Diaconis & R. Graham & B. Sturmfels: Primitive partition identities, Paul Erdös is 80. Vol. II, Janos Bolyai Society, Budapest (1995) 1–20Google Scholar
- 3.R. E. Gomory: Some polyhedra related to combinatorial problems, Linear Algebra and its Applications 2 (1969) 451–558Google Scholar
- 4.J. E. Graver: On the foundations of linear and integer programming I, Mathematical Programming 8 (1975) 207–226Google Scholar
- 5.R. Kannan: Solution of the Frobenius problem and its generalizations, Manuscript (1991)Google Scholar
- 6.E. S. Selmer: On the linear diophantine problem of Frobenius, Journal für reine und angewandte Mathematik 293/294 (1977), 1–17Google Scholar
- 7.R. R. Thomas: Gröbner basis methods for integer programming, PhD. Dissertation, Cornell University, (1994)Google Scholar
- 8.R. Urbaniak & R. Weismantel & G. Ziegler: A variant of Buchberger's algorithm for integer programming, SIAM J. Discrete Math., Vol 1, No 10, (1997) 96–108Google Scholar