English

Efficient sequential and parallel algorithms for multistage stochastic integer programming using proximity

Data Structures and Algorithms 2020-12-23 v1 Computational Complexity Optimization and Control

Abstract

We consider the problem of solving integer programs of the form min{cx  ⁣: Ax=b,x0}\min \{\,c^\intercal x\ \colon\ Ax=b, x\geq 0\}, where AA is a multistage stochastic matrix in the following sense: the primal treedepth of AA is bounded by a parameter dd, which means that the columns of AA can be organized into a rooted forest of depth at most dd so that columns not bound by the ancestor/descendant relation in the forest do not have non-zero entries in the same row. We give an algorithm that solves this problem in fixed-parameter time f(d,A)nlogO(2d)nf(d,\|A\|_{\infty})\cdot n\log^{O(2^d)} n, where ff is a computable function and nn is the number of rows of AA. The algorithm works in the strong model, where the running time only measures unit arithmetic operations on the input numbers and does not depend on their bitlength. This is the first fpt algorithm for multistage stochastic integer programming to achieve almost linear running time in the strong sense. For the case of two-stage stochastic integer programs, our algorithm works in time 2(2A)O(r(r+s))nlogO(rs)n2^{(2\|A\|_\infty)^{O(r(r+s))}}\cdot n\log^{O(rs)} n. The algorithm can be also parallelized: we give an implementation in the PRAM model that achieves running time f(d,A)logO(2d)nf(d,\|A\|_{\infty})\cdot \log^{O(2^d)} n using nn processors. The main conceptual ingredient in our algorithms is a new proximity result for multistage stochastic integer programs. We prove that if we consider an integer program PP, say with a constraint matrix AA, then for every optimum solution to the linear relaxation of PP there exists an optimum (integral) solution to PP that lies, in the \ell_{\infty}-norm, within distance bounded by a function of A\|A\|_{\infty} and the primal treedepth of AA. On the way to achieve this result, we prove a generalization and considerable improvement of a structural result of Klein for multistage stochastic integer programs.

Keywords

Cite

@article{arxiv.2012.11742,
  title  = {Efficient sequential and parallel algorithms for multistage stochastic integer programming using proximity},
  author = {Jana Cslovjecsek and Friedrich Eisenbrand and Michał Pilipczuk and Moritz Venzin and Robert Weismantel},
  journal= {arXiv preprint arXiv:2012.11742},
  year   = {2020}
}

Comments

23 pages, 2 figures