The UGC hardness threshold of the Lp grothendieck problem

Guy Kindler, Assaf Naor, Gideon Schechtman

Research output: Contribution to journalArticle

Abstract

For p ≥ 2 we consider the problem of, given an n x n matrix A = (a ij) whose diagonal entries vanish, approximating in polynomial time the number Optp(A):= max { σi, j=1 n aijxixj: (x1,....,xn) ∈ ℝn ∧ (σi=1 n|x i|p)1/p≤1}.When p = 2 this is simply the problem of computing the maximum eigenvalue of A, whereas for p = ∞ (actually it suffices to take p≈ log n) it is the Grothendieck problem on the complete graph, which was shown to have a 0(log n) approximation algorithm in Nemirovski et al. [Nemirovski, A., C. Roos, T. Terlaky. 1999. On maximization of quadratic form over intersection of ellipsoids with common center. Math. Program. Ser. A 86(3) 463-473], Megretski [Megretski, A. 2001. Relaxations of quadratic programs in operator theory and system analysis. Systems, Approximation, Singular Integral Operators, and Related Topics (Bordeaux, 2000), Vol. 129. Operator Theory Advances and Applications. Birkhäuser, Basel, 365-392], Charikar and Wirth [Charikar, M., A. Wirth. 2004. Maximizing quadratic programs: Extending Grothendieck's inequality. Proc. 45th Annual Sympos. Foundations Comput. Sci., IEEE Computer Society, 54-60] and was used in the work of Charikar and Wirth noted above, to design the best known algorithm for the problem of computing the maximum correlation in correlation clustering. Thus the problem of approximating Optp(A) interpolates between the spectral (p = 2) case and the correlation clustering (p =∞ ) case. From a physics point of view this problem corresponds to computing the ground states of spin glasses in a hard-wall potential well. We design a polynomial time algorithm which, given p ≥ 2 and an n x n matrix A = (aij) with zeros on the diagonal, computes Optp(A) up to a factor p/e + 30logp. On the other hand, assuming the unique games conjecture (UGC) we show that it is NP-hard to approximate Optp(A) up to a factor smaller than p/e + 1 4. Hence as p --∞ the UGC-hardness threshold for computing Optp(A) is exactly (p/e)(1 + o(1)).

Original languageEnglish (US)
Pages (from-to)267-283
Number of pages17
JournalMathematics of Operations Research
Volume35
Issue number2
DOIs
StatePublished - May 2010

Fingerprint

Hardness
Polynomials
Game
Spin glass
Approximation algorithms
Ground state
Quadratic Program
Operator Theory
Physics
Computing
Systems analysis
Clustering
Singular Integral Operator
Potential Well
Spin Glass
Ellipsoid
Systems Analysis
Quadratic form
Complete Graph
Polynomial-time Algorithm

Keywords

  • Approximation algorithms
  • Quadratic programming
  • Unique games hardness

ASJC Scopus subject areas

  • Mathematics(all)
  • Computer Science Applications
  • Management Science and Operations Research

Cite this

The UGC hardness threshold of the Lp grothendieck problem. / Kindler, Guy; Naor, Assaf; Schechtman, Gideon.

In: Mathematics of Operations Research, Vol. 35, No. 2, 05.2010, p. 267-283.

Research output: Contribution to journalArticle

Kindler, Guy ; Naor, Assaf ; Schechtman, Gideon. / The UGC hardness threshold of the Lp grothendieck problem. In: Mathematics of Operations Research. 2010 ; Vol. 35, No. 2. pp. 267-283.
@article{cb3bd8b8e3264feb94f32ae8ab8069a5,
title = "The UGC hardness threshold of the Lp grothendieck problem",
abstract = "For p ≥ 2 we consider the problem of, given an n x n matrix A = (a ij) whose diagonal entries vanish, approximating in polynomial time the number Optp(A):= max { σi, j=1 n aijxixj: (x1,....,xn) ∈ ℝn ∧ (σi=1 n|x i|p)1/p≤1}.When p = 2 this is simply the problem of computing the maximum eigenvalue of A, whereas for p = ∞ (actually it suffices to take p≈ log n) it is the Grothendieck problem on the complete graph, which was shown to have a 0(log n) approximation algorithm in Nemirovski et al. [Nemirovski, A., C. Roos, T. Terlaky. 1999. On maximization of quadratic form over intersection of ellipsoids with common center. Math. Program. Ser. A 86(3) 463-473], Megretski [Megretski, A. 2001. Relaxations of quadratic programs in operator theory and system analysis. Systems, Approximation, Singular Integral Operators, and Related Topics (Bordeaux, 2000), Vol. 129. Operator Theory Advances and Applications. Birkh{\"a}user, Basel, 365-392], Charikar and Wirth [Charikar, M., A. Wirth. 2004. Maximizing quadratic programs: Extending Grothendieck's inequality. Proc. 45th Annual Sympos. Foundations Comput. Sci., IEEE Computer Society, 54-60] and was used in the work of Charikar and Wirth noted above, to design the best known algorithm for the problem of computing the maximum correlation in correlation clustering. Thus the problem of approximating Optp(A) interpolates between the spectral (p = 2) case and the correlation clustering (p =∞ ) case. From a physics point of view this problem corresponds to computing the ground states of spin glasses in a hard-wall potential well. We design a polynomial time algorithm which, given p ≥ 2 and an n x n matrix A = (aij) with zeros on the diagonal, computes Optp(A) up to a factor p/e + 30logp. On the other hand, assuming the unique games conjecture (UGC) we show that it is NP-hard to approximate Optp(A) up to a factor smaller than p/e + 1 4. Hence as p --∞ the UGC-hardness threshold for computing Optp(A) is exactly (p/e)(1 + o(1)).",
keywords = "Approximation algorithms, Quadratic programming, Unique games hardness",
author = "Guy Kindler and Assaf Naor and Gideon Schechtman",
year = "2010",
month = "5",
doi = "10.1287/moor.1090.0425",
language = "English (US)",
volume = "35",
pages = "267--283",
journal = "Mathematics of Operations Research",
issn = "0364-765X",
publisher = "INFORMS Inst.for Operations Res.and the Management Sciences",
number = "2",

}

TY - JOUR

T1 - The UGC hardness threshold of the Lp grothendieck problem

AU - Kindler, Guy

AU - Naor, Assaf

AU - Schechtman, Gideon

PY - 2010/5

Y1 - 2010/5

N2 - For p ≥ 2 we consider the problem of, given an n x n matrix A = (a ij) whose diagonal entries vanish, approximating in polynomial time the number Optp(A):= max { σi, j=1 n aijxixj: (x1,....,xn) ∈ ℝn ∧ (σi=1 n|x i|p)1/p≤1}.When p = 2 this is simply the problem of computing the maximum eigenvalue of A, whereas for p = ∞ (actually it suffices to take p≈ log n) it is the Grothendieck problem on the complete graph, which was shown to have a 0(log n) approximation algorithm in Nemirovski et al. [Nemirovski, A., C. Roos, T. Terlaky. 1999. On maximization of quadratic form over intersection of ellipsoids with common center. Math. Program. Ser. A 86(3) 463-473], Megretski [Megretski, A. 2001. Relaxations of quadratic programs in operator theory and system analysis. Systems, Approximation, Singular Integral Operators, and Related Topics (Bordeaux, 2000), Vol. 129. Operator Theory Advances and Applications. Birkhäuser, Basel, 365-392], Charikar and Wirth [Charikar, M., A. Wirth. 2004. Maximizing quadratic programs: Extending Grothendieck's inequality. Proc. 45th Annual Sympos. Foundations Comput. Sci., IEEE Computer Society, 54-60] and was used in the work of Charikar and Wirth noted above, to design the best known algorithm for the problem of computing the maximum correlation in correlation clustering. Thus the problem of approximating Optp(A) interpolates between the spectral (p = 2) case and the correlation clustering (p =∞ ) case. From a physics point of view this problem corresponds to computing the ground states of spin glasses in a hard-wall potential well. We design a polynomial time algorithm which, given p ≥ 2 and an n x n matrix A = (aij) with zeros on the diagonal, computes Optp(A) up to a factor p/e + 30logp. On the other hand, assuming the unique games conjecture (UGC) we show that it is NP-hard to approximate Optp(A) up to a factor smaller than p/e + 1 4. Hence as p --∞ the UGC-hardness threshold for computing Optp(A) is exactly (p/e)(1 + o(1)).

AB - For p ≥ 2 we consider the problem of, given an n x n matrix A = (a ij) whose diagonal entries vanish, approximating in polynomial time the number Optp(A):= max { σi, j=1 n aijxixj: (x1,....,xn) ∈ ℝn ∧ (σi=1 n|x i|p)1/p≤1}.When p = 2 this is simply the problem of computing the maximum eigenvalue of A, whereas for p = ∞ (actually it suffices to take p≈ log n) it is the Grothendieck problem on the complete graph, which was shown to have a 0(log n) approximation algorithm in Nemirovski et al. [Nemirovski, A., C. Roos, T. Terlaky. 1999. On maximization of quadratic form over intersection of ellipsoids with common center. Math. Program. Ser. A 86(3) 463-473], Megretski [Megretski, A. 2001. Relaxations of quadratic programs in operator theory and system analysis. Systems, Approximation, Singular Integral Operators, and Related Topics (Bordeaux, 2000), Vol. 129. Operator Theory Advances and Applications. Birkhäuser, Basel, 365-392], Charikar and Wirth [Charikar, M., A. Wirth. 2004. Maximizing quadratic programs: Extending Grothendieck's inequality. Proc. 45th Annual Sympos. Foundations Comput. Sci., IEEE Computer Society, 54-60] and was used in the work of Charikar and Wirth noted above, to design the best known algorithm for the problem of computing the maximum correlation in correlation clustering. Thus the problem of approximating Optp(A) interpolates between the spectral (p = 2) case and the correlation clustering (p =∞ ) case. From a physics point of view this problem corresponds to computing the ground states of spin glasses in a hard-wall potential well. We design a polynomial time algorithm which, given p ≥ 2 and an n x n matrix A = (aij) with zeros on the diagonal, computes Optp(A) up to a factor p/e + 30logp. On the other hand, assuming the unique games conjecture (UGC) we show that it is NP-hard to approximate Optp(A) up to a factor smaller than p/e + 1 4. Hence as p --∞ the UGC-hardness threshold for computing Optp(A) is exactly (p/e)(1 + o(1)).

KW - Approximation algorithms

KW - Quadratic programming

KW - Unique games hardness

UR - http://www.scopus.com/inward/record.url?scp=77953100811&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77953100811&partnerID=8YFLogxK

U2 - 10.1287/moor.1090.0425

DO - 10.1287/moor.1090.0425

M3 - Article

VL - 35

SP - 267

EP - 283

JO - Mathematics of Operations Research

JF - Mathematics of Operations Research

SN - 0364-765X

IS - 2

ER -