English

A scaling-invariant algorithm for linear programming whose running time depends only on the constraint matrix

Optimization and Control 2023-03-16 v4 Data Structures and Algorithms

Abstract

Following the breakthrough work of Tardos in the bit-complexity model, Vavasis and Ye gave the first exact algorithm for linear programming in the real model of computation with running time depending only on the constraint matrix. For solving a linear program (LP) maxcx,Ax=b,x0,ARm×n\max\, c^\top x,\: Ax = b,\: x \geq 0,\: A \in \mathbb{R}^{m \times n}, Vavasis and Ye developed a primal-dual interior point method using a 'layered least squares' (LLS) step, and showed that O(n3.5log(χˉA+n))O(n^{3.5} \log (\bar{\chi}_A+n)) iterations suffice to solve (LP) exactly, where χˉA\bar{\chi}_A is a condition measure controlling the size of solutions to linear systems related to AA. Monteiro and Tsuchiya, noting that the central path is invariant under rescalings of the columns of AA and cc, asked whether there exists an LP algorithm depending instead on the measure χˉA\bar{\chi}^*_A, defined as the minimum χˉAD\bar{\chi}_{AD} value achievable by a column rescaling ADAD of AA, and gave strong evidence that this should be the case. We resolve this open question affirmatively. Our first main contribution is an O(m2n2+n3)O(m^2 n^2 + n^3) time algorithm which works on the linear matroid of AA to compute a nearly optimal diagonal rescaling DD satisfying χˉADn(χˉ)3\bar{\chi}_{AD} \leq n(\bar{\chi}^*)^3. This algorithm also allows us to approximate the value of χˉA\bar{\chi}_A up to a factor n(χˉ)2n (\bar{\chi}^*)^2. As our second main contribution, we develop a scaling invariant LLS algorithm, together with a refined potential function based analysis for LLS algorithms in general. With this analysis, we derive an improved O(n2.5lognlog(χˉA+n))O(n^{2.5} \log n\log (\bar{\chi}^*_A+n)) iteration bound for optimally solving (LP) using our algorithm. The same argument also yields a factor n/lognn/\log n improvement on the iteration complexity bound of the original Vavasis-Ye algorithm.

Keywords

Cite

@article{arxiv.1912.06252,
  title  = {A scaling-invariant algorithm for linear programming whose running time depends only on the constraint matrix},
  author = {Daniel Dadush and Sophie Huiberts and Bento Natura and László A. Végh},
  journal= {arXiv preprint arXiv:1912.06252},
  year   = {2023}
}
R2 v1 2026-06-23T12:44:41.324Z