+ All Categories
Home > Documents > Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011...

Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011...

Date post: 19-Jan-2016
Category:
Upload: myron-nash
View: 214 times
Download: 0 times
Share this document with a friend
21
Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton University & CCI Boaz Barak MSR New England U
Transcript
Page 1: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

Subexponential Algorithms for Unique Games and Related Problems

Approximation Algorithms, June 2011

David SteurerMSR New England

Sanjeev Arora Princeton University & CCI

Boaz BarakMSR New England

U

Page 2: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

Subexponential Algorithms for Unique Games and Related Problems

David SteurerMSR New England

Sanjeev Arora Princeton University & CCI

Boaz BarakMSR New England

Rounding Semidefinite ProgrammingHierarchies via Global Correlation

David SteurerMSR New England

Prasad RaghavendraGeorgia Tech

Boaz BarakMSR New England

Page 3: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

UNIQUE GAMESInput: list of constraints of form

Goal: satisfy as many constraints as possible

[๐‘˜ ][๐‘˜ ]

๐‘ฅ ๐‘—๐‘ฅ๐‘–

Page 4: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

UNIQUE GAMESInput: list of constraints of form

Goal: satisfy as many constraints as possible

Input: UNIQUE GAMES instance with (say)

Goal: Distinguish two cases

YES: more than of constraints satisfiableNO: less than of constraints

satisfiable

Unique Games Conjecture (UGC) [Khotโ€™02]

For every , the following is NP-hard:

UG (๐œ€)

Page 5: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

Implications of UGCFor many basic optimization problems, it is NP-hard to beat current algorithms (based on simple LP or SDP relaxations)

Examples:

VERTEX COVER [Khot-Regevโ€™03], MAX CUT [Khot-Kindler-Mossel-Oโ€™Donnellโ€™04,

Mossel-Oโ€™Donnell-Oleszkiewiczโ€™05],every MAX CSP [Raghavendraโ€™08], โ€ฆ

Page 6: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

Implications of UGCFor many basic optimization problems, it is NP-hard to beat current algorithms (based on simple LP or SDP relaxations)

Unique Games Barrier

Example: -approximation for MAX CUTat least as hard as

UNIQUE GAMES is common barrier for improving current algorithms of

many basic problems

Goemansโ€“Williamson

bound for MAX CUT

Page 7: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

Subexponential Algorithm for Unique Games

Input: UNIQUE GAMES instance with alphabet size ksuch that of constraints are satisfiable,

Output:assignment satisfying of constraints Time:

in time

Time vs Approximation Trade-off

Page 8: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

Analog of UGC with subconstant (say ) is false (*)(contrast: subconstant hardness for LABEL COVER [Moshkovitz-Razโ€™08])

NP-hardness reduction for must have blow-up (*)

rules out certain classes of reductions for proving UGC

(*) assuming 3-SAT does not have subexponential algorithms,

UGC-based hardness does not rule out subexponential algorithms, Possibility: -time algorithm for MAX CUT() ?

Subexponential Algorithm for Unique Games

in time

Consequences

Page 9: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

poly (๐‘›) exp (๐‘›)

2-SAT

MAX 3-SAT()MAX CUT()

3-SAT (*)

FACTORING

exp (๐‘›1 /2)exp (๐‘›1 /3)exp (๐‘›๐œ€ 1/3 )

UG (๐œ€ )MAX 3-SAT()

LABEL COVER()

[Moshkovitz-Razโ€™08+ Hรฅstadโ€™97]MAX CUT()?

(*) assuming Exponential Time Hypothesis [Impagliazzo-Paturi-Zaneโ€™01]( 3-SAT has no algorithm )

Subexponential Algorithm for Unique Games

in time

GRAPH ISOMORPHISM

Page 10: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

Subexponential Algorithm for Unique Games

in time

here: via semidefinite programming hierarchies

Page 11: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

What we want:

jointly distributed random variables over

UNIQUE GAMESInput: list of constraints of form

Goal: satisfy as many constraints as possible

Pr (๐‘‹ ๐‘–โˆ’ ๐‘‹ ๐‘—โ‰ก๐‘ )โ‰ฅ1โˆ’๐œ€ for typical constraint

Page 12: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

= indicator of

p.s.d. for all with -

same marginal for in distributions and

-local

Goal: produce global random variables

for most constraints

Here: iterative procedure for [Arora-Barak-S.โ€™10+ Barak-Raghavendra-S.โ€™11]

jointly distributed random variables over

Pr (๐‘‹ ๐‘–โˆ’ ๐‘‹ ๐‘—โ‰ก๐‘ )โ‰ฅ1โˆ’๐œ€ for typical constraint

distributions over for all with

time

consistency

positive semidefiniteness

๐‘†๐‘‡

๐‘†โˆฉ๐‘‡

-level SDP hierarchy:What we want:

Page 13: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

Components of iterative procedure

Rounding

Conditioning

sample variables independently according to their marginals

pick a vertex and sample

condition on sample for

Partitioning

find vertex subset

break dependence between and

enough if constraint graph has few significant eigenvalues

[BRSโ€™11]

general frameworkfor rounding

SDP hierarchies

Page 14: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

Important fact:

can approximate by Gram matrix of unit vectors (tensoring trick)

Corr (๐‘‹ ๐‘– , ๐‘‹ ๐‘—)(Pairwise) Correlation

max๐‘

โˆ‘๐‘Ž

|Cov ( ๐‘‹ ๐‘–๐‘Ž , ๐‘‹ ๐‘— (๐‘Ž+๐‘ ) )|

โˆ‘๐‘Ž

(๐‘‰๐‘Ž๐‘Ÿ (๐‘‹ ๐‘–๐‘Ž )โˆ’๐‘‰๐‘Ž๐‘Ÿ (๐‘‹ ๐‘–๐‘Ž|๐‘‹ ๐‘—ยฟ )decrease in variance when conditioning on

statistical distance between and

Rounding Conditioning Partitioning

similar to mutual information

Page 15: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

Rounding Conditioning Partitioning

Page 16: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

sample variables independently according to their marginals

If then independent samplingsatisfies constraint with probability

Rounding fails ๐„๐‘–โˆผ ๐‘—Corr (๐‘‹ ๐‘– ,๐‘‹ ๐‘— )>1โˆ’๐‘‚ (๐œ€)

Local Correlation(over edges of constraint graph)

Rounding Conditioning Partitioning

statistical distance between independent and correlated

sampling

Page 17: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

pick a vertex and sample

condition on sample for

computationally expensive(level level )

condition on vertex only if

can condition times on such vertices

Conditioning fails

Global Correlation(over random vertex pairs)

Rounding Conditioning Partitioning

Issue:

Idea:

measures decrease in variance when conditioning

on

Page 18: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

find vertex subset

break dependence between and

destroy correlation for constraints between and

๐‘‰ โˆ–๐‘†

๐‘†

Rounding Conditioning Partitioning

Issue:

Wanted:set with small expansion & small cardinality

Page 19: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

For endpoints of random path of length

๐„๐‘–โˆผ๐‘ก ๐‘—Corr (๐‘‹ ๐‘– ,๐‘‹ ๐‘— )>(1โˆ’๐‘‚ (๐œ€) )๐‘กโ‰ซ๐‘›โˆ’๐›ฝ

random walk stuck for steps on fraction of vertices vertex set with and expansion

A vertex is cut in at most partitioning steps break only edges

Rounding Conditioning Partitioning

๐‘‰ โˆ–๐‘†

๐‘†

fails only if local correlation high

fails only if global correlation low

Correlation Propagation

is Grammatrix of unit

vectors

find vertex subset

break dependence between and

Page 20: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

For general 2-CSP:

PTAS if constraint graph is random (degree alphabet)

QPTAS if constraint graph is hypercontractive very good expander

for small sets

[Barak-Raghavendra-S.โ€™11]

Subsequent work: [Arora-Geโ€™11]

better 3-COLORING approximation on some graph families

Independent work:

approximation schemes for quadratic integer programming with p.s.d. objective & few relevant eigenvalues

More SDP-hierarchy algorithms

[Guruswami-Sinopโ€™11]

Page 21: Subexponential Algorithms for Unique Games and Related Problems Approximation Algorithms, June 2011 David Steurer MSR New England Sanjeev Arora Princeton.

Open Questions

Example: -approximation for SPARSEST CUT in time ?

How many large eigenvalues can a small-set expander have?

Is Boolean noise graph the worst case? (large eigenvalues)

Thank you!Questions?

What else can be done in subexponential time?

Towards settling the Unique Games Conjecture

Better approximations for MAX CUT or VERTEX COVER on general instances?

No: small set expander with large eigenvalues

[Barak-Gopalan-Hรฅstad-Meka-Raghavendra-S.โ€™11]


Recommended