English

$\ell_1$-sparsity Approximation Bounds for Packing Integer Programs

Data Structures and Algorithms 2019-02-26 v1

Abstract

We consider approximation algorithms for packing integer programs (PIPs) of the form max{c,x:Axb,x{0,1}n}\max\{\langle c, x\rangle : Ax \le b, x \in \{0,1\}^n\} where cc, AA, and bb are nonnegative. We let W=mini,jbi/Ai,jW = \min_{i,j} b_i / A_{i,j} denote the width of AA which is at least 11. Previous work by Bansal et al. \cite{bansal-sparse} obtained an Ω(1Δ01/W)\Omega(\frac{1}{\Delta_0^{1/\lfloor W \rfloor}})-approximation ratio where Δ0\Delta_0 is the maximum number of nonzeroes in any column of AA (in other words the 0\ell_0-column sparsity of AA). They raised the question of obtaining approximation ratios based on the 1\ell_1-column sparsity of AA (denoted by Δ1\Delta_1) which can be much smaller than Δ0\Delta_0. Motivated by recent work on covering integer programs (CIPs) \cite{cq,chs-16} we show that simple algorithms based on randomized rounding followed by alteration, similar to those of Bansal et al. \cite{bansal-sparse} (but with a twist), yield approximation ratios for PIPs based on Δ1\Delta_1. First, following an integrality gap example from \cite{bansal-sparse}, we observe that the case of W=1W=1 is as hard as maximum independent set even when Δ12\Delta_1 \le 2. In sharp contrast to this negative result, as soon as width is strictly larger than one, we obtain positive results via the natural LP relaxation. For PIPs with width W=1+ϵW = 1 + \epsilon where ϵ(0,1]\epsilon \in (0,1], we obtain an Ω(ϵ2/Δ1)\Omega(\epsilon^2/\Delta_1)-approximation. In the large width regime, when W2W \ge 2, we obtain an Ω((11+Δ1/W)1/(W1))\Omega((\frac{1}{1 + \Delta_1/W})^{1/(W-1)})-approximation. We also obtain a (1ϵ)(1-\epsilon)-approximation when W=Ω(log(Δ1/ϵ)ϵ2)W = \Omega(\frac{\log (\Delta_1/\epsilon)}{\epsilon^2}).

Keywords

Cite

@article{arxiv.1902.08698,
  title  = {$\ell_1$-sparsity Approximation Bounds for Packing Integer Programs},
  author = {Chandra Chekuri and Kent Quanrud and Manuel R. Torres},
  journal= {arXiv preprint arXiv:1902.08698},
  year   = {2019}
}

Comments

To appear in IPCO 2019