Inapproximability of MAX-CUT Khot,Kindler,Mossel and O ’ Donnell Moshe Ben Nehemia June 05.

Slides:



Advertisements
Similar presentations
Hardness of Reconstructing Multivariate Polynomials. Parikshit Gopalan U. Washington Parikshit Gopalan U. Washington Subhash Khot NYU/Gatech Rishi Saket.
Advertisements

On allocations that maximize fairness Uriel Feige Microsoft Research and Weizmann Institute.
Vertex cover might be hard to approximate within 2 - ε Subhash Khot, Oded Regev Slides by: Ofer Neiman.
A threshold of ln(n) for approximating set cover By Uriel Feige Lecturer: Ariel Procaccia.
Shortest Vector In A Lattice is NP-Hard to approximate
Subhash Khot IAS Elchanan Mossel UC Berkeley Guy Kindler DIMACS Ryan O’Donnell IAS.
Inapproximability of Hypergraph Vertex-Cover. A k-uniform hypergraph H= : V – a set of vertices E - a collection of k-element subsets of V Example: k=3.
Hardness of Approximating Multicut S. Chawla, R. Krauthgamer, R. Kumar, Y. Rabani, D. Sivakumar (2005) Presented by Adin Rosenberg.
Approximate List- Decoding and Hardness Amplification Valentine Kabanets (SFU) joint work with Russell Impagliazzo and Ragesh Jaiswal (UCSD)
Circuit and Communication Complexity. Karchmer – Wigderson Games Given The communication game G f : Alice getss.t. f(x)=1 Bob getss.t. f(y)=0 Goal: Find.
Gillat Kol joint work with Ran Raz Locally Testable Codes Analogues to the Unique Games Conjecture Do Not Exist.
Dana Moshkovitz MIT Joint work with Subhash Khot, NYU 1.
The Max-Cut problem: Election recounts? Majority vs. Electoral College? 7812.
C&O 355 Lecture 23 N. Harvey TexPoint fonts used in EMF. Read the TexPoint manual before you delete this box.: A A A A A A A A A A.
On the Unique Games Conjecture Subhash Khot Georgia Inst. Of Technology. At FOCS 2005.
Approximation Algorithms Chapter 14: Rounding Applied to Set Cover.
Inapproximability Seminar – 2005 David Arnon  March 3, 2005 Some Optimal Inapproximability Results Johan Håstad Royal Institute of Technology, Sweden.
Gibbs sampler - simple properties It’s not hard to show that this MC chain is aperiodic. Often is reversible distribution. If in addition the chain is.
MaxClique Inapproximability Seminar on HARDNESS OF APPROXIMATION PROBLEMS by Dr. Irit Dinur Presented by Rica Gonen.
Constraint Satisfaction over a Non-Boolean Domain Approximation Algorithms and Unique Games Hardness Venkatesan Guruswami Prasad Raghavendra University.
Parallel Scheduling of Complex DAGs under Uncertainty Grzegorz Malewicz.
Semi-Definite Algorithm for Max-CUT Ran Berenfeld May 10,2005.
Introduction to PCP and Hardness of Approximation Dana Moshkovitz Princeton University and The Institute for Advanced Study 1.
A 3-Query PCP over integers a.k.a Solving Sparse Linear Systems Prasad Raghavendra Venkatesan Guruswami.
1/17 Optimal Long Test with One Free Bit Nikhil Bansal (IBM) Subhash Khot (NYU)
Complexity ©D.Moshkovits 1 Hardness of Approximation.
Inapproximability from different hardness assumptions Prahladh Harsha TIFR 2011 School on Approximability.
Dictator tests and Hardness of approximating Max-Cut-Gain Ryan O’Donnell Carnegie Mellon (includes joint work with Subhash Khot of Georgia Tech)
The number of edge-disjoint transitive triples in a tournament.
Fast FAST By Noga Alon, Daniel Lokshtanov And Saket Saurabh Presentation by Gil Einziger.
Complexity 16-1 Complexity Andrei Bulatov Non-Approximability.
Approximation Algoirthms: Semidefinite Programming Lecture 19: Mar 22.
1 Tight Hardness Results for Some Approximation Problems [Raz,Håstad,...] Adi Akavia Dana Moshkovitz & Ricky Rosen S. Safra.
Semidefinite Programming
Vol.1: Geometry Subhash Khot IAS Elchanan Mossel UC Berkeley Guy Kindler DIMACS Ryan O’Donnell IAS.
1 Tight Hardness Results for Some Approximation Problems [mostly Håstad] Adi Akavia Dana Moshkovitz S. Safra.
Distributed Combinatorial Optimization
Complexity 1 Hardness of Approximation. Complexity 2 Introduction Objectives: –To show several approximation problems are NP-hard Overview: –Reminder:
On the hardness of approximating Sparsest-Cut and Multicut Shuchi Chawla, Robert Krauthgamer, Ravi Kumar, Yuval Rabani, D. Sivakumar.
1 Introduction to Approximation Algorithms Lecture 15: Mar 5.
1 Joint work with Shmuel Safra. 2 Motivation 3 Motivation.
1 Hardness Result for MAX-3SAT This lecture is given by: Limor Ben Efraim.
Dana Moshkovitz, MIT Joint work with Subhash Khot, NYU.
Outline Introduction The hardness result The approximation algorithm.
Fractional decompositions of dense hypergraphs Raphael Yuster University of Haifa.
Correlation testing for affine invariant properties on Shachar Lovett Institute for Advanced Study Joint with Hamed Hatami (McGill)
HARDNESS OF APPROXIMATIONS. Gap Introducing Reduction For simplicity we assume that we are always reducing from SAT(or any other NP- hard problem). Let.
1 Introduction to Approximation Algorithms. 2 NP-completeness Do your best then.
C&O 355 Mathematical Programming Fall 2010 Lecture 16 N. Harvey TexPoint fonts used in EMF. Read the TexPoint manual before you delete this box.: A A A.
1/19 Minimizing weighted completion time with precedence constraints Nikhil Bansal (IBM) Subhash Khot (NYU)
Non-Approximability Results. Summary -Gap technique -Examples: MINIMUM GRAPH COLORING, MINIMUM TSP, MINIMUM BIN PACKING -The PCP theorem -Application:
Linear Program Set Cover. Given a universe U of n elements, a collection of subsets of U, S = {S 1,…, S k }, and a cost function c: S → Q +. Find a minimum.
C&O 355 Lecture 24 N. Harvey TexPoint fonts used in EMF. Read the TexPoint manual before you delete this box.: AA A A A A A A A A.
Unique Games Approximation Amit Weinstein Complexity Seminar, Fall 2006 Based on: “Near Optimal Algorithms for Unique Games" by M. Charikar, K. Makarychev,
Non-LP-Based Approximation Algorithms Fabrizio Grandoni IDSIA
Approximation Algorithms Greedy Strategies. I hear, I forget. I learn, I remember. I do, I understand! 2 Max and Min  min f is equivalent to max –f.
CHAPTER SIX T HE P ROBABILISTIC M ETHOD M1 Zhang Cong 2011/Nov/28.
Richard Anderson Lecture 26 NP-Completeness
Richard Anderson Lecture 26 NP-Completeness
Lecture 18: Uniformity Testing Monotonicity Testing
Computability and Complexity
Subhash Khot Dept of Computer Science NYU-Courant & Georgia Tech
Introduction to PCP and Hardness of Approximation
Approximation Algorithms
Parameterized Complexity of Even Set (and others)
The Complexity of Approximation
Switching Lemmas and Proof Complexity
Instructor: Aaron Roth
Locality In Distributed Graph Algorithms
Small Set Expansion in The Johnson Graph
Presentation transcript:

Inapproximability of MAX-CUT Khot,Kindler,Mossel and O ’ Donnell Moshe Ben Nehemia June 05

Main Result  It is NP-Hard problem to approximate MAX-CUT to within a factor  is the approximation ratio achieved by the algorithm of Goemans & Williamson.  The result follows from:  1. Unique Games conjecture  2. Majority is Stablest Theorem

Hardness of Approximation  History:  Bellare & Goldreich & Sudan :It is NP Hard to approximate MAX-CUT within factor higher than 83/84  Hasted improved the result to 16/17  Today: closing the gap …

Introduction  MAX-CUT: Given a Graph G =(V,E), find a partition C=(V 1, V 2 ) that maximize:  Unique Label Cover: Given a bi-partite graph with left side vertices- V,right side W, and edges- E each edge have a constraint bijection The goal: assign each vertex a label which satisfy the constraint.

Unique Games Conjecture: For any there exist a constant Such that it is NP-hard to distinguish whether the Unique Label Cover problem with label set in size M has optimum at least or at most

Some defintions  Let be an arbitrary boolean function  The influence of x i on f  Let x be a uniformly random string in :E[X]=0 and form y by flipping each bit with prob The noise stability of f for a noise rate is:

 The Correlation between x,y is define to be: E[XY] = 2 Pr[X=Y]-1  Let x be a uniformly random string in y be -correlated copy :i.e. pick each bit independently s.t. The noise correlation of f with parameter is:

 Result[60 ’ ] :

 Fix then for any there is a small enough s.t. if is any function satisfying : Then: The Majority is Stablest Theorem

On the Geometry of MAX-CUT  The Goemans-Williamson algorithm:  Embedding the graph in the unit sphere of R n :  The embedding is selected s.t. this sum is maximize A cut in G is obtained by choosing a random hyperplane through the origin. And this sum bounds from above the size of the maximal cut

On the Geometry of MAX-CUT The probability that vertics u,v lie on opposite sides of the cut is: So the expected weight is

On the Geometry of MAX-CUT  So to get:  Set the approximation ratio to:

Reminder  The Long Code:  The codeword encoding the message is by the truth table of the “ dictator ” function:

Technical Background  The Bonami Beckner operator Proposition: Let and then:

Technical Background Proposition: Let then for every Proof: Define: And : And using the Parseval identity we get the proposition

Technical Background  Let and let The k-degree influence of coordinate i on f is defined by: Proposition: The “ Majority is Stablest ” Theorem remains true if we change the assumption to

Reverse version of the “Majority is Stablest”  Fix then for any there is a small enough s.t. if is any function satisfying : Then:

Reverse version of the “Majority is Stablest”  Proof:  Take such f, and define:  Now g holds:  And now apply the original Theorem

Reduction from Unique LC to MAX-CUT Notations: denote the string and xy the coordinatewise product of x and y  Lemma 1: Completeness  If ULC have OPT then MAX-CUT have cut  Lemma 2: Soundness  If ULC have OPT then MAX-CUT have cut at most

Reduction from Unique LC to MAX-CUT Unique Label Cover WV v W’W’ w MAX-CUT j J’J’ i {-1,1} M

Reduction from Unique LC to MAX-CUT  The Reduction:  Pick a vertex at random and 2 of its neighbors:  Let and be the constrains for those edges  Let f,g be the supposed Long Codes of the labels  Pick at random  Pick by choosing each coordinate independently to be 1 with probability and -1 with prob. Edge in Cut iff

Reduction from Unique LC to MAX-CUT  Completeness Assume that the LC instance has a labeling which satisfies fraction of the edges. now encode these labels via Long Code with prob both the edges are satisfied by the labeling Denote the label of v,w,w ’ by i,j,j ’

Reduction from Unique LC to MAX-CUT  Completeness note that: Now f,g are the Long Codes of j,j ’, so: The two bits are unequal iff and that happens with prob. hence the completeness :

Reduction from Unique LC to MAX-CUT  Soundness – The Proof Strategy if the max-cut bigger than we ’ ll be able to “ decode ” the “ Long Code “ and create a labeling which satisfy significant fraction of the edges in the LC problem, and get a contradiction by choosing small enough.

Reduction from Unique LC to MAX-CUT From the Fourier Transform:

Reduction from Unique LC to MAX-CUT  The expectation over x vanishes unless and then s,s ’ have the same size. Because: We got: Because of for at least v in V ( “ good ” v) We have

Reduction from Unique LC to MAX-CUT Define Now:

Reduction from Unique LC to MAX-CUT  Now,from the “ Majority is stablest ” theorem:  We conclude that h has at least one coordinate j s.t. label the vertex v with j And we have:

Reduction from Unique LC to MAX-CUT  From the above equation we have that for at least fraction of neighbors w of v we have Define And so, Because we got that

Reduction from Unique LC to MAX-CUT  Now,if we label each vertex w in W by random element from Cand[w], then among the “ good ” vertices v at least satisfied. or among the edges, and that yields the contradiction

Restriction to the cube  For 2 vectors x,y on the cube,define the weight of (x,y) to be: Pr[X=x and Y=y] where X,Y are -correlated elements on the cube  A cut C defines a boolean function on the cube f c.  The size of C is exactly  The expected size of the maximal cut is  But, the “ dictator cut “ has size The “ Majority is stablest ” conjecture claim that these are the only cases.

The strategy of the proof  correctness A legal code word (dictator func) has Pr[acc]=  Soundness let be a func far from being a “ long code ” from the Majority is Stablest we get so the test pass with prob: