English

Optimization Problems with Diseconomies of Scale via Decoupling

Data Structures and Algorithms 2015-01-26 v2 Probability

Abstract

We present a new framework for solving optimization problems with a diseconomy of scale. In such problems, our goal is to minimize the cost of resources used to perform a certain task. The cost of resources grows superlinearly, as xqx^q, q1q\ge 1, with the amount xx of resources used. We define a novel linear programming relaxation for such problems, and then show that the integrality gap of the relaxation is AqA_q, where AqA_q is the qq-th moment of the Poisson random variable with parameter 1. Using our framework, we obtain approximation algorithms for the Minimum Energy Efficient Routing, Minimum Degree Balanced Spanning Tree, Load Balancing on Unrelated Parallel Machines, and Unrelated Parallel Machine Scheduling with Nonlinear Functions of Completion Times problems. Our analysis relies on the decoupling inequality for nonnegative random variables. The inequality states that i=1nXiqCqi=1nYiq,\big \|\sum_{i=1}^n X_i\big\|_{q} \leq C_q \,\big \|\sum_{i=1}^n Y_i\big\|_{q}, where XiX_i are independent nonnegative random variables, YiY_i are possibly dependent nonnegative random variable, and each YiY_i has the same distribution as XiX_i. The inequality was proved by de la Pe\~na in 1990. De la Pe\~na, Ibragimov, and Sharakhmetov (2003) showed that Cq2C_q\leq 2 for q(1,2)q\in (1,2) and CqAq1/qC_q\leq A_q^{1/q} for q2q\geq 2. We show that the optimal constant is Cq=Aq1/qC_q=A_q^{1/q} for any q1q\geq 1. We then prove a more general inequality: For every convex function φ\varphi, E[φ(i=1nXi)]E[φ(Pi=1nYi)],\mathbb{E}[\varphi\Big(\sum_{i=1}^n X_i\Big)]\leq \mathbb{E}[\varphi\Big(P\sum_{i=1}^n Y_i\Big)], and, for every concave function ψ\psi, E[ψ(i=1nXi)]E[ψ(Pi=1nYi)],\mathbb{E}[\psi\Big(\sum_{i=1}^n X_i\Big)] \geq \mathbb{E}[\psi\Big(P\sum_{i=1}^n Y_i\Big)], where PP is a Poisson random variable with parameter 1 independent of the random variables YiY_i.

Keywords

Cite

@article{arxiv.1404.3248,
  title  = {Optimization Problems with Diseconomies of Scale via Decoupling},
  author = {Konstantin Makarychev and Maxim Sviridenko},
  journal= {arXiv preprint arXiv:1404.3248},
  year   = {2015}
}
R2 v1 2026-06-22T03:49:12.643Z