Constraint Satisfaction over a Non-Boolean Domain Approximation Algorithms and Unique Games Hardness Venkatesan Guruswami Prasad Raghavendra University of Washington Seattle, WA Constraint Satisfaction Problem A Classic Example : Max-3-SAT ( x1 x 2 x3 )( x2 x3 x5 )( x 2 x3 x5 )( x5 x4 x1 ) Equivalently the largest fraction of clauses Given a 3-SAT formula, Find an assignment to the variables that satisfies the maximum number of clauses. Constraint Satisfaction Problem General Definition : Example : Max-3-SAT Domain : {0,1,.. q-1} Predicates : {P1, P2 , P3 … Pr} Domain : Predicates : Pi : [q]k -> {0,1} Arity : Maximum number of variables per constraint (k) {0,1} P1(x,y,z) = x ѵ y ѵ z Arity = 3 GOAL : Find an assignment satisfying maximum fraction of constraints Approximability Most Max-CSP problems are NP-hard to solve exactly. Different Max-CSP problems are approximable to varying ratios. Max-2-SAT : 0.94 [Lempel-LivnatZwick],[Austrin] Max-3-SAT : 7/8 [Karloff-Zwick],[Hastad] 3-XOR : ½ [Hastad] Max-Cut : 0.878 [Goemans-Williamson], [Khot-Kindler-MosselO’donnel] Refined Question : Question : Among all Max-CSP problems over Which Max-CSP is the hardest to domain [q] ={0,..q-1}, and arity k, approximate? which is the hardest to approximate? Clearly, the problems become harder as domain size or the arity grows PCP Motivation Completeness(C) : If SAT formula is satisfiable, there is a proof that verifier accepts with probability C Soundness (S): For an unsatisfiable formula, no proof is accepted with probability more than S ( x1 x2 x3 )( x2 x3 x5 )( x2 x3 x5 )( x5 x4 x1 ) Probabilistically Checkable Proof (A string over alphabet {0,1,..q-1}) Random bits Verifier ACCEPT/REJECT PCP Motivation What is the best possible gap between completeness (c) and soundness (s) for a PCP verifier that makes k queries over an alphabet [q] = {0,1,..q-1} ? Among all Max-CSP problems over domain [q] ={0,..q-1}, and arity k, which is the hardest to approximate? Boolean CSPs Hardness: For every k, there is a boolean CSP of arity k, which is NPhard to approximate better than : [Samorodnitsky-Trevisan 2000] simplified by [Hastad-Wigderson] [Engebresten-Holmerin] [Charikar-Makarychev-Makarychev] Every boolean CSP of arity k, can be approximated to a factor : 2 2k 2k Assuming Unique Games Conjecture [Samorodnitsky-Trevisan 2006] Algorithm: 22 k 2k [Hast] [Trevisan] Random Assignment 2k 2k 0.88k 2k k / log k 2k 2 2k 1 2k This Work : Non-Boolean CSPs UG Hardness: Assuming Unique Games Conjecture, For every k, and a prime number q, there is a CSP of arity k over the domain [q] ={0,1,2,..q-1}, which is NP-hard to approximate better than Algorithm: The algorithm of [Charikar-MakarychevMakarychev] can be extended to non-boolean domains. Every CSP of arity k over the domain [q] ={0,1,2,..q-1} can be approximated to a factor 2 q k qk k / q7 k q Related Work “Optimal approximation algorithms and hardness results for every CSP, assuming Unique Games Conjecture.” [Raghavendra 08] – “Every” so applies to the hardest CSPs too. – Does not give explicit example of hardest CSP, nor the explicit value of the approximation ratio. [Austrin-Mossel 08] “Assuming Unique Games conjecture, For every prime power q, and k, it is NP-hard to q(q 1)k approximate a certain CSP over [q] to a factor > ” k q – Independent work using entirely different techniques(invariance principle) – Show a more general result, that yields a criteria for Approximation Resistance of a predicate. Techniques • We extend the proof techniques of [Samorodnitsky-Trevisan 2006] to non- boolean domains. • To this end, we – Define a subspace linearity test. – Show a technical lemma relating the success probability of a function F to the Gower’s norm of F (similar to the standard proof relating the number of multidimensional arithmetic progressions to the Gower’s norm) • Along the way, we make some minor simplifications to [Samorodnitsky-Trevisan 2006]. – (Remove the need for common influences) Proof Overview Dictatorship Testing Problem Given a function F : [q]R [q], • Make at most k queries to F •Based on values of F, Output ACCEPT or REJECT. Distinguish between the following two cases : F is a dictator function F(x1 ,… xR) = xi Pr[ACCEPT ] = Completeness F is far from every dictator function (No influential coordinate) Pr[ACCEPT ] = Soundness Goal : Achieve maximum gap between Completeness and Soundness UG Hardness Proofs Test For the restDictatorship of the talk, weOver shall focus F:[q]R -> [q] functions on Dictatorship Completeness = C Soundness = S Testing. # of queries = k Using [Khot-Kindler-Mossel-O’Donnell] reduction. UG Hardness Result: Assuming Unique Games Conjecture, it is NPhard to approximate a CSP over [q] with arity k to ratio better than C/S Testing Dictatorships by Testing Linearity [Samorodnitsky-Trevisan 2006] Fix {0,1} : field on 2 elements k = 2d Given a function F : {0,1}R -> {0,1} • Pick a random affine subspace A of dimension d. • Test if F agrees with some affine linear function on the subspace A. Random Assignment : There are 2d+1 different Every dictator affine linear functions on A. F(x1 , x2 ,.. xR ) = xi 2d is a linear function There are 2 possible functions on A.space over vector {0,1}R So a random function satisfies the test with probability 2 d 1 2k 2 2d 2k Cubes Gower’s Norm x+y2 x x+y1 For F : {0,1}R -> {0,1}, let f(x) = (-1)F(x) . dth Gowers Norm Ud(f) = E[ product of f over C] Expectation over random ddimensional subcubes C in {0,1}R x+y1+y2 x x+y3+ y2 x+y2 x+y1+y2+y3 x+y1+y2 x+y3 x x+y1 x+y1+y3 x+y1 d-dimensional cube spanned by {x,y1 ,y2 ,.. yd } is C x yi | S {1,..d } iS Gower’s Norm More Formally, Intuitively, the dth Gower’s norm measures the correlation of the function f with degree d-1 polynomials. Testing Dictatorships by Testing Linearity [Samorodnitsky-Trevisan 2006] Lemma : If F : {0,1}R -> {0,1} passes the test with probability 2 k then f = (-1)F has high dth Gowers k 2 Norm. (k=2d) R -> {-1,1} has Lemma : If a balanced function f : {0,1} Using Noise sensitivity, high dth Gowers Norm, then it has There are an onlyinfluential a FEW coordinate (k=2d) influential coordinates. Theorem : If a balanced function F : {0,1}R -> {0,1} 2k passes the test with probability k then it has an 2 influential coordinate Extending to Larger domains Fix [q]: field on q elements(q is a prime). k = qd Given a function F : [q]R -> [q] • Pick a random affine subspace A of dimension d. • Test if F agrees with some affine linear function on the subspace A. Replace 2 by q in the [SamorodnitskyTrevisan] dictatorship test. The Difficulty Lemma : If F : {0,1}R -> {0,1} passes the test with probability 2 k then f = (-1)F has high dth Gowers k 2 Norm. (k=2d) Over {0,1}R, Subcube = Affine subspace. Testing linearity over a random affine subspace, can be easily related to expectation over a random cube. Over [q]R , Subcube ≠ Affine subspace. (2R points) (qR points) Success probability of a function F : {0,1}R -> {0,1}, is related to : let f(x) = (-1)F(x) . E[ product of f over A] Multidimensional Progressions x x+y1 x+(q-1)y2 x+2y1 x+(q-2)y1x+(q-1)y1 x+(q-1)y2+y1 x+(q-1)y2+2y1x+(q-1)y2+ (q-2)yx+(q-1)y 1 2+ (q-1)y1 Expectation over random d-dimensional affine subspace A in [q]R (Affine subspaces are like multidimensional arithmetic progressions) E[ product of f over C] Expectation over random dq-dimensional subcubes C in [q]R x+(q-2)y2 x+(q-2)y2+y1 x+(q-2)y2+2y1x+(q-2)y2+ (q-2)yx+(q-2)y 2+ (q-1)y1 1 x+2y2 x+2y2+y1 x+2y2+2y1 x+2y2+ (q-2)y1 x+y2 x+y2+y1 x+y2+2y1 x+y2+ (q-2)y1 x+y2+ (q-1)y1 x x+y1 x+2y1 x+ (q-2)y1 x+2y2+ (q-1)y1 x+ (q-1)y1 Alternate Lemma Lemma : If F : [q]R -> [q] passes the test with 2 q probability k then f = (-1)F has high dqth Gower’s k Norm. (k=qd) q •d-dimensional affine subspace test relates to the dqth Gower’s norm •The proof is technical and involves repeated use of the Cauchy-Schwartz inequality. • A special case of a more general result by [Green-Tao][Gowers-Wolf], where they define “Cauchy-Schwartz Complexity” of a set of linear forms. Open Questions CSPs with Perfect Completeness: Which CSP is hardest to approximate, under the promise that the input instance is completely satisfiable? Approximation Resistance: Characterize CSPs for which the best approximation achievable is given by a random assignment. Thank You Unique Games A Special Case x-y = 11 (mod 17) x-z = 13 (mod 17) … …. z-w = 15(mod 17) E2LIN mod p Given a set of linear equations of the form: Xi – Xj = cij mod p Find a solution that satisfies the maximum number of equations. Unique Games Conjecture [Khot 02] An Equivalent Version [Khot-Kindler-Mossel-O’Donnell] For every ε> 0, the following problem is NP-hard for large enough prime p Given a E2LIN mod p system, distinguish between: • There is an assignment satisfying 1-ε fraction of the equations. • No assignment satisfies more than ε fraction of equations.