English

Sum-of-squares hierarchies for binary polynomial optimization

Optimization and Control 2022-01-20 v3

Abstract

We consider the sum-of-squares hierarchy of approximations for the problem of minimizing a polynomial ff over the boolean hypercube Bn={0,1}n\mathbb{B}^{n}=\{0,1\}^n. This hierarchy provides for each integer rNr \in \mathbb{N} a lower bound f(r)f_{(r)} on the minimum fminf_{\min} of ff, given by the largest scalar λ\lambda for which the polynomial fλf - \lambda is a sum-of-squares on Bn\mathbb{B}^{n} with degree at most 2r2r. We analyze the quality of these bounds by estimating the worst-case error fminf(r)f_{\min} - f_{(r)} in terms of the least roots of the Krawtchouk polynomials. As a consequence, for fixed t[0,1/2]t \in [0, 1/2], we can show that this worst-case error in the regime rtnr \approx t \cdot n is of the order 1/2t(1t)1/2 - \sqrt{t(1-t)} as nn tends to \infty. Our proof combines classical Fourier analysis on Bn\mathbb{B}^{n} with the polynomial kernel technique and existing results on the extremal roots of Krawtchouk polynomials. This link to roots of orthogonal polynomials relies on a connection between the hierarchy of lower bounds f(r)f_{(r)} and another hierarchy of upper bounds f(r)f^{(r)}, for which we are also able to establish the same error analysis. Our analysis extends to the minimization of a polynomial over the qq-ary cube (Z/qZ)n(\mathbb{Z}/q\mathbb{Z})^{n}.

Keywords

Cite

@article{arxiv.2011.04027,
  title  = {Sum-of-squares hierarchies for binary polynomial optimization},
  author = {Lucas Slot and Monique Laurent},
  journal= {arXiv preprint arXiv:2011.04027},
  year   = {2022}
}

Comments

31 pages, 1 figure. Version 3: Changed structure of the paper. Added clarifying statements, shortened several proofs. Fixed minor mistakes/typos

R2 v1 2026-06-23T19:59:37.870Z