SDP gaps and UGC-hardness for MAXCUTGAIN

Subhash Khot, Ryan O'Donneil

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Given a graph with maximum cut of (fractional) size c, the Goemans-Williamson [GW95] semidefinite programming (SDP) algorithm is guaranteed to find a cut of size .878 &· c. However this guarantee becomes trivial when c is near 1/2, since a random cut has expected size 1/2. Recently, Charikar and Worth [CW04] (analyzing an algorithm of Feige and Langberg [FL01]) showed that given a graph with maximum cut 1/2 + ε, one can find a cut of size 1/2 + ω (ε / log(1/ε)). The main contribution of our paper is twofold: 1. We give a natural 1/2 + ε vs. 1/2 + O(ε/ log(1/ε)) SDP gap for MAXCUT in Gaussian space. This shows that the SDP-rounding algorithm of Charikar-Worth is essentially best possible. Further, the "s-linear rounding functions" used in [CW04, FL01] arise as optimizers in our analysis, somewhat confirming a suggestion of [FL01]. 2. We show how this SDP gap can be translated into a Long Code test with the same parameters. This implies that beating the Charikar-Worth guarantee with any efficient algorithm is NP-hard, assuming the Unique Games Conjecture (UGC) [Kho02]. We view this result as essentially settling the approximability of MAXCUT, assuming UGC. Building on (1) we show how "randomness reduction" on related SDP gaps for the QUADRATICPRO-GRAMMlNG programming problem lets us make the Ω(log(1/ε)) gap as large as Ω(log n) for n-vertex graphs. In addition to optimally answering an open question of [AMMN06], this technique may prove useful for other SDP gap problems. Finally, illustrating the generality of our technique in (2), we also show how to translate Reeds's [Ree93] SDP gap for the Grothendieck Inequality into a UGC-hardness result for computing the ∥·∥ ∞→1 norm of a matrix.

Original languageEnglish (US)
Title of host publication47th Annual IEEE Symposium on Foundations of Computer Science, FOCS 2006
Pages217-226
Number of pages10
DOIs
StatePublished - 2006
Event47th Annual IEEE Symposium on Foundations of Computer Science, FOCS 2006 - Berkeley, CA, United States
Duration: Oct 21 2006Oct 24 2006

Other

Other47th Annual IEEE Symposium on Foundations of Computer Science, FOCS 2006
CountryUnited States
CityBerkeley, CA
Period10/21/0610/24/06

Fingerprint

Hardness

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Khot, S., & O'Donneil, R. (2006). SDP gaps and UGC-hardness for MAXCUTGAIN. In 47th Annual IEEE Symposium on Foundations of Computer Science, FOCS 2006 (pp. 217-226). [4031358] https://doi.org/10.1109/FOCS.2006.67

SDP gaps and UGC-hardness for MAXCUTGAIN. / Khot, Subhash; O'Donneil, Ryan.

47th Annual IEEE Symposium on Foundations of Computer Science, FOCS 2006. 2006. p. 217-226 4031358.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Khot, S & O'Donneil, R 2006, SDP gaps and UGC-hardness for MAXCUTGAIN. in 47th Annual IEEE Symposium on Foundations of Computer Science, FOCS 2006., 4031358, pp. 217-226, 47th Annual IEEE Symposium on Foundations of Computer Science, FOCS 2006, Berkeley, CA, United States, 10/21/06. https://doi.org/10.1109/FOCS.2006.67
Khot S, O'Donneil R. SDP gaps and UGC-hardness for MAXCUTGAIN. In 47th Annual IEEE Symposium on Foundations of Computer Science, FOCS 2006. 2006. p. 217-226. 4031358 https://doi.org/10.1109/FOCS.2006.67
Khot, Subhash ; O'Donneil, Ryan. / SDP gaps and UGC-hardness for MAXCUTGAIN. 47th Annual IEEE Symposium on Foundations of Computer Science, FOCS 2006. 2006. pp. 217-226
@inproceedings{530c7a7fc6684b6b8f1d117332517373,
title = "SDP gaps and UGC-hardness for MAXCUTGAIN",
abstract = "Given a graph with maximum cut of (fractional) size c, the Goemans-Williamson [GW95] semidefinite programming (SDP) algorithm is guaranteed to find a cut of size .878 &· c. However this guarantee becomes trivial when c is near 1/2, since a random cut has expected size 1/2. Recently, Charikar and Worth [CW04] (analyzing an algorithm of Feige and Langberg [FL01]) showed that given a graph with maximum cut 1/2 + ε, one can find a cut of size 1/2 + ω (ε / log(1/ε)). The main contribution of our paper is twofold: 1. We give a natural 1/2 + ε vs. 1/2 + O(ε/ log(1/ε)) SDP gap for MAXCUT in Gaussian space. This shows that the SDP-rounding algorithm of Charikar-Worth is essentially best possible. Further, the {"}s-linear rounding functions{"} used in [CW04, FL01] arise as optimizers in our analysis, somewhat confirming a suggestion of [FL01]. 2. We show how this SDP gap can be translated into a Long Code test with the same parameters. This implies that beating the Charikar-Worth guarantee with any efficient algorithm is NP-hard, assuming the Unique Games Conjecture (UGC) [Kho02]. We view this result as essentially settling the approximability of MAXCUT, assuming UGC. Building on (1) we show how {"}randomness reduction{"} on related SDP gaps for the QUADRATICPRO-GRAMMlNG programming problem lets us make the Ω(log(1/ε)) gap as large as Ω(log n) for n-vertex graphs. In addition to optimally answering an open question of [AMMN06], this technique may prove useful for other SDP gap problems. Finally, illustrating the generality of our technique in (2), we also show how to translate Reeds's [Ree93] SDP gap for the Grothendieck Inequality into a UGC-hardness result for computing the ∥·∥ ∞→1 norm of a matrix.",
author = "Subhash Khot and Ryan O'Donneil",
year = "2006",
doi = "10.1109/FOCS.2006.67",
language = "English (US)",
isbn = "0769527205",
pages = "217--226",
booktitle = "47th Annual IEEE Symposium on Foundations of Computer Science, FOCS 2006",

}

TY - GEN

T1 - SDP gaps and UGC-hardness for MAXCUTGAIN

AU - Khot, Subhash

AU - O'Donneil, Ryan

PY - 2006

Y1 - 2006

N2 - Given a graph with maximum cut of (fractional) size c, the Goemans-Williamson [GW95] semidefinite programming (SDP) algorithm is guaranteed to find a cut of size .878 &· c. However this guarantee becomes trivial when c is near 1/2, since a random cut has expected size 1/2. Recently, Charikar and Worth [CW04] (analyzing an algorithm of Feige and Langberg [FL01]) showed that given a graph with maximum cut 1/2 + ε, one can find a cut of size 1/2 + ω (ε / log(1/ε)). The main contribution of our paper is twofold: 1. We give a natural 1/2 + ε vs. 1/2 + O(ε/ log(1/ε)) SDP gap for MAXCUT in Gaussian space. This shows that the SDP-rounding algorithm of Charikar-Worth is essentially best possible. Further, the "s-linear rounding functions" used in [CW04, FL01] arise as optimizers in our analysis, somewhat confirming a suggestion of [FL01]. 2. We show how this SDP gap can be translated into a Long Code test with the same parameters. This implies that beating the Charikar-Worth guarantee with any efficient algorithm is NP-hard, assuming the Unique Games Conjecture (UGC) [Kho02]. We view this result as essentially settling the approximability of MAXCUT, assuming UGC. Building on (1) we show how "randomness reduction" on related SDP gaps for the QUADRATICPRO-GRAMMlNG programming problem lets us make the Ω(log(1/ε)) gap as large as Ω(log n) for n-vertex graphs. In addition to optimally answering an open question of [AMMN06], this technique may prove useful for other SDP gap problems. Finally, illustrating the generality of our technique in (2), we also show how to translate Reeds's [Ree93] SDP gap for the Grothendieck Inequality into a UGC-hardness result for computing the ∥·∥ ∞→1 norm of a matrix.

AB - Given a graph with maximum cut of (fractional) size c, the Goemans-Williamson [GW95] semidefinite programming (SDP) algorithm is guaranteed to find a cut of size .878 &· c. However this guarantee becomes trivial when c is near 1/2, since a random cut has expected size 1/2. Recently, Charikar and Worth [CW04] (analyzing an algorithm of Feige and Langberg [FL01]) showed that given a graph with maximum cut 1/2 + ε, one can find a cut of size 1/2 + ω (ε / log(1/ε)). The main contribution of our paper is twofold: 1. We give a natural 1/2 + ε vs. 1/2 + O(ε/ log(1/ε)) SDP gap for MAXCUT in Gaussian space. This shows that the SDP-rounding algorithm of Charikar-Worth is essentially best possible. Further, the "s-linear rounding functions" used in [CW04, FL01] arise as optimizers in our analysis, somewhat confirming a suggestion of [FL01]. 2. We show how this SDP gap can be translated into a Long Code test with the same parameters. This implies that beating the Charikar-Worth guarantee with any efficient algorithm is NP-hard, assuming the Unique Games Conjecture (UGC) [Kho02]. We view this result as essentially settling the approximability of MAXCUT, assuming UGC. Building on (1) we show how "randomness reduction" on related SDP gaps for the QUADRATICPRO-GRAMMlNG programming problem lets us make the Ω(log(1/ε)) gap as large as Ω(log n) for n-vertex graphs. In addition to optimally answering an open question of [AMMN06], this technique may prove useful for other SDP gap problems. Finally, illustrating the generality of our technique in (2), we also show how to translate Reeds's [Ree93] SDP gap for the Grothendieck Inequality into a UGC-hardness result for computing the ∥·∥ ∞→1 norm of a matrix.

UR - http://www.scopus.com/inward/record.url?scp=35448938376&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=35448938376&partnerID=8YFLogxK

U2 - 10.1109/FOCS.2006.67

DO - 10.1109/FOCS.2006.67

M3 - Conference contribution

SN - 0769527205

SN - 9780769527208

SP - 217

EP - 226

BT - 47th Annual IEEE Symposium on Foundations of Computer Science, FOCS 2006

ER -