Bypassing UGC from Some Optimal Geometric Inapproximability Results

The Unique Games Conjecture (UGC) has emerged in recent years as the starting point for several optimal inapproximability results. While for <i>none</i> of these results a reverse reduction to Unique Games is known, the assumption of bijective projections in the Label Cover instance nevertheless seems critical in these proofs. In this work, we bypass the need for UGC assumption in inapproximability results for two geometric problems, obtaining a tight NP-hardness result in each case. The first problem, known as <i>L<sub>p</sub></i> Subspace Approximation, is a generalization of the classic least squares regression problem. Here, the input consists of a set of points <i>X</i> = {α<sub>1</sub>, … , α<sub>m</sub>} ⊆ R<i><sub>n</sub></i> and a parameter <i>k</i> (possibly depending on <i>n</i>). The goal is to find a subspace <i>H</i> of <i>R</i><i><sub>n</sub></i> of dimension <i>k</i> that minimizes the ℓ<sub><i>p</i></sub> norm of the Euclidean distances to the points in <i>X</i>. For <i>p</i> = 2, <i>k</i> = <i>n</i> − 1, this reduces to the least squares regression problem, while for <i>p</i> = ∞, <i>k</i> = 0 it reduces to the problem of finding a ball of minimum radius enclosing all the points. We show that for any fixed <i>p</i> ∈ (2, ∞), and for <i>k</i> = <i>n</i> − 1, it is NP-hard to approximate this problem to within a factor of γ<sub><i>p</i></sub> − ε for constant ε > 0, where γ<sub><i>p</i></sub> is the <i>p</i>th norm of a standard Gaussian random variable. This matches the γ<sub><i>p</i></sub> approximation algorithm obtained by Deshpande, Tulsiani, and Vishnoi who also showed the same hardness result under the UGC. The second problem we study is the related <i>L<sub>p</sub></i> Quadratic Grothendieck Maximization Problem, considered by Kindler, Naor, and Schechtman. Here, the input is a multilinear quadratic form ∑<sub><i>n</i></sub><sub><i>i</i>, <i>j</i> = 1</sub><i>a<sub>ij</sub>x<sub>i</sub>x<sub>j</sub></i> and the goal is to maximize the quadratic form over the ℓ<sub><i>p</i></sub> unit ball, namely, all <i>x</i> with ∑<sub><i>n</i></sub><sub><i>i</i> = 1</sub>|<i>x<sub>i</sub></i>|<sub><i>p</i></sub> ⩽ 1. The problem is polynomial time solvable for <i>p</i> = 2. We show that for any constant <i>p</i> ∈ (2, ∞), it is NP-hard to approximate the quadratic form to within a factor of γ<sub>2</sub><sub><i>p</i></sub> − ε for any ε > 0. The same hardness factor was shown under the UGC by Kindler et al. We also obtain a γ<sub>2</sub><sub><i>p</i></sub>-approximation algorithm for the problem using the convex relaxation of the problem defined by Kindler et al. A γ<sub>2</sub><sub><i>p</i></sub> approximation algorithm has also been independently obtained by Naor and Schechtman. These are the <i>first</i> approximation thresholds, proven under P ≠ NP, that involve the Gaussian random variable in a fundamental way. Note that the problem statements themselves do not explicitly involve the Gaussian distribution.

[1]  Guy Kindler,et al.  The UGC hardness threshold of the ℓp Grothendieck problem , 2008, SODA '08.

[2]  Eran Halperin,et al.  HAPLOFREQ-Estimating Haplotype Frequencies Efficiently , 2006, J. Comput. Biol..

[3]  Elchanan Mossel Gaussian Bounds for Noise Correlation of Functions , 2007, FOCS 2007.

[4]  Prasad Raghavendra,et al.  Optimal algorithms and inapproximability results for every CSP? , 2008, STOC.

[5]  Elchanan Mossel,et al.  Gaussian Bounds for Noise Correlation of Functions and Tight Analysis of Long Codes , 2008, 2008 49th Annual IEEE Symposium on Foundations of Computer Science.

[6]  N. Alon,et al.  Quadratic forms on graphs , 2006 .

[7]  Kasturi R. Varadarajan,et al.  Sampling-based dimension reduction for subspace approximation , 2007, STOC '07.

[8]  Ryan O'Donnell,et al.  Optimal Inapproximability Results for MAX-CUT and Other 2-Variable CSPs? , 2007, SIAM J. Comput..

[9]  Y. Nesterov Global quadratic optimization via conic relaxation , 1998 .

[10]  Subhash Khot On the power of unique 2-prover 1-round games , 2002, STOC '02.

[11]  Kasturi R. Varadarajan,et al.  Efficient Subspace Approximation Algorithms , 2007, Discrete & Computational Geometry.

[12]  Nisheeth K. Vishnoi,et al.  Algorithms and hardness for subspace approximation , 2009, SODA '11.

[13]  Guy Kindler,et al.  On non-approximability for quadratic programs , 2005, 46th Annual IEEE Symposium on Foundations of Computer Science (FOCS'05).

[14]  Moses Charikar,et al.  Near-optimal algorithms for unique games , 2006, STOC '06.

[15]  Luca Trevisan,et al.  Approximation algorithms for unique games , 2005, 46th Annual IEEE Symposium on Foundations of Computer Science (FOCS'05).

[16]  David P. Woodruff,et al.  Coresets and sketches for high dimensional subspace approximation problems , 2010, SODA '10.

[17]  Sanjeev Arora,et al.  Subexponential Algorithms for Unique Games and Related Problems , 2010, 2010 IEEE 51st Annual Symposium on Foundations of Computer Science.

[18]  Carsten Lund,et al.  Proof verification and the hardness of approximation problems , 1998, JACM.

[19]  Juliane Jung Approximation And Complexity In Numerical Optimization Continuous And Discrete Problems , 2016 .

[20]  Subhash Khot,et al.  Vertex cover might be hard to approximate to within 2-epsilon , 2008, J. Comput. Syst. Sci..

[21]  Noga Alon,et al.  Approximating the cut-norm via Grothendieck's inequality , 2004, STOC '04.

[22]  Tamás Terlaky,et al.  On maximization of quadratic form over intersection of ellipsoids with common center , 1999, Math. Program..

[23]  Prasad Raghavendra,et al.  Beating the Random Ordering Is Hard: Every Ordering CSP Is Approximation Resistant , 2011, SIAM J. Comput..

[24]  Nisheeth K. Vishnoi,et al.  The unique games conjecture, integrality gap for cut problems and embeddability of negative type metrics into l/sub 1/ , 2005, 46th Annual IEEE Symposium on Foundations of Computer Science (FOCS'05).

[25]  Gene H. Golub,et al.  Matrix computations , 1983 .

[26]  Subhash Khot,et al.  On hardness of learning intersection of two halfspaces , 2008, STOC '08.

[27]  Venkatesan Guruswami,et al.  Lasserre Hierarchy, Higher Eigenvalues, and Approximation Schemes for Graph Partitioning and Quadratic Integer Programming with PSD Objectives , 2011, 2011 IEEE 52nd Annual Symposium on Foundations of Computer Science.

[28]  Immanuel M. Bomze,et al.  Regularity versus Degeneracy in Dynamics, Games, and Optimization: A Unified Approach to Different Aspects , 2002, SIAM Rev..

[29]  Srinivasan Venkatesh,et al.  Approximating the Radii of Point Sets , 2007, SIAM J. Comput..

[30]  Subhash Khot,et al.  A 3-query non-adaptive PCP with perfect completeness , 2006, 21st Annual IEEE Conference on Computational Complexity (CCC'06).

[31]  Nisheeth K. Vishnoi,et al.  The Unique Games Conjecture, Integrality Gap for Cut Problems and Embeddability of Negative Type Metrics into l1 , 2005, FOCS.

[32]  Prasad Raghavendra,et al.  Agnostic Learning of Monomials by Halfspaces Is Hard , 2009, 2009 50th Annual IEEE Symposium on Foundations of Computer Science.

[33]  Nisheeth K. Vishnoi,et al.  Unique games on expanding constraint graphs are easy: extended abstract , 2008, STOC '08.

[34]  Subhash Khot,et al.  Hardness results for coloring 3-colorable 3-uniform hypergraphs , 2002, The 43rd Annual IEEE Symposium on Foundations of Computer Science, 2002. Proceedings..

[35]  Nisheeth K. Vishnoi,et al.  Integrality gaps for sparsest cut and minimum linear arrangement problems , 2006, STOC '06.

[36]  A. Megretski Relaxations of Quadratic Programs in Operator Theory and System Analysis , 2001 .

[37]  Subhash Khot,et al.  On the hardness of learning intersections of two halfspaces , 2011, J. Comput. Syst. Sci..

[38]  Peter Gritzmann,et al.  Inapproximability of some Geometric and Quadratic Optimization Problems , 2000 .

[39]  Uri Zwick,et al.  Approximation algorithms for constraint satisfaction problems involving at most three variables per constraint , 1998, SODA '98.

[40]  Noga Alon,et al.  Approximating the Cut-Norm via Grothendieck's Inequality , 2006, SIAM J. Comput..

[41]  Sanjeev Arora,et al.  Probabilistic checking of proofs: a new characterization of NP , 1998, JACM.

[42]  Subhash Khot,et al.  Optimal Long Code Test with One Free Bit , 2009, 2009 50th Annual IEEE Symposium on Foundations of Computer Science.

[43]  Prasad Raghavendra,et al.  Rounding Semidefinite Programming Hierarchies via Global Correlation , 2011, 2011 IEEE 52nd Annual Symposium on Foundations of Computer Science.

[44]  Prasad Raghavendra,et al.  Integrality Gaps for Strong SDP Relaxations of UNIQUE GAMES , 2009, 2009 50th Annual IEEE Symposium on Foundations of Computer Science.

[45]  Moses Charikar,et al.  Maximizing quadratic programs: extending Grothendieck's inequality , 2004, 45th Annual IEEE Symposium on Foundations of Computer Science.

[46]  Sariel Har-Peled,et al.  Projective clustering in high dimensions using core-sets , 2002, SCG '02.

[47]  Subhash Khot,et al.  Hardness of Reconstructing Multivariate Polynomials over Finite Fields , 2007, FOCS.

[48]  Ran Raz A Parallel Repetition Theorem , 1998, SIAM J. Comput..

[49]  Anupam Gupta,et al.  Approximating unique games , 2006, SODA '06.

[50]  Alexandra Kolla Spectral Algorithms for Unique Games , 2010, 2010 IEEE 25th Annual Conference on Computational Complexity.

[51]  Subhash Khot,et al.  Inapproximability of Hypergraph Vertex Cover and Applications to Scheduling Problems , 2010, ICALP.