Tight Size-Degree Bounds for Sums-of-Squares Proofs

AbstractWe exhibit families of 4-CNF formulas over n variables that have sums-of-squares (SOS) proofs of unsatisfiability of degree (a.k.a. rank) d but require SOS proofs of size $${n^{\Omega{(d)}}}$$nΩ(d) for values of d = d(n) from constant all the way up to $${n^{\delta}}$$nδ for some universal constant $${\delta}$$δ. This shows that the $${{n^{{\rm O}{(d)}}}}$$nO(d) running time obtained by using the Lasserre semidefinite programming relaxations to find degree-d SOS proofs is optimal up to constant factors in the exponent. We establish this result by combining NP-reductions expressible as low-degree SOS derivations with the idea of relativizing CNF formulas in Krajíček (Arch Math Log 43(4):427–441, 2004) and Dantchev & Riis (Proceedings of the 17th international workshop on computer science logic (CSL ’03), 2003) and then applying a restriction argument as in Atserias et al. (J Symb Log 80(2):450–476, 2015; ACM Trans Comput Log 17:19:1–19:30, 2016). This yields a generic method of amplifying SOS degree lower bounds to size lower bounds and also generalizes the approach used in Atserias et al. (2016) to obtain size lower bounds for the proof systems resolution, polynomial calculus, and Sherali–Adams from lower bounds on width, degree, and rank, respectively.

[1]  Søren Riis,et al.  On Relativisation and Complexity Gap , 2003, CSL.

[2]  Albert Atserias,et al.  Narrow Proofs May Be Maximally Long , 2014, Computational Complexity Conference.

[3]  G. Stengle.  A nullstellensatz and a positivstellensatz in semialgebraic geometry , 1974 .

[4]  N. Z. Shor.  An approach to obtaining global extremums in polynomial mathematical programming problems , 1987 .

[5]  P. Parrilo.  Structured semidefinite programs and semialgebraic geometry methods in robustness and optimization , 2000 .

[6]  Yuan Zhou,et al.  Hypercontractivity, sum-of-squares proofs, and their applications , 2012, STOC '12.

[7]  Olaf Beyersdorff,et al.  Parameterized Complexity of DPLL Search Procedures , 2011, SAT.

[8]  References , 1971 .

[9]  Jean B. Lasserre,et al.  An Explicit Exact SDP Relaxation for Nonlinear 0-1 Programs , 2001, IPCO.

[10]  Dima Grigoriev,et al.  Linear lower bound on degrees of Positivstellensatz calculus proofs for the parity , 2001, Theor. Comput. Sci..

[11]  Toniann Pitassi,et al.  Exponential Lower Bounds and Integrality Gaps for Tree-Like Lovász-Schrijver Procedures , 2009, SIAM J. Comput..

[12]  Alexander A. Razborov,et al.  Parameterized Bounded-Depth Frege Is not Optimal , 2011, TOCT.

[13]  J. Krivine,et al.  Anneaux préordonnés , 1964 .

[14]  Stefan S. Dantchev.  Relativisation Provides Natural Separations for Resolution-Based Proof Systems , 2006, CSR.

[15]  Toniann Pitassi,et al.  Lower Bounds for Lovász-Schrijver Systems and Beyond Follow from Multiparty Communication Complexity , 2005, ICALP.

[16]  Prasad Raghavendra,et al.  On the Power of Symmetric LP and SDP Relaxations , 2014, 2014 IEEE 29th Conference on Computational Complexity (CCC).

[17]  Vojtech Rödl,et al.  The complexity of proving that a graph is Ramsey , 2017, Comb..

[18]  Madhur Tulsiani.  CSP gaps and reductions in the lasserre hierarchy , 2009, STOC '09.

[19]  Albert Atserias,et al.  Lower Bounds for DNF-refutations of a Relativized Weak Pigeonhole Principle , 2013, 2013 IEEE Conference on Computational Complexity.

[20]  Dima Grigoriev,et al.  Exponential Lower Bound for Static Semi-algebraic Proofs , 2002, ICALP.

[21]  Jean-Pierre Bourguignon,et al.  Mathematische Annalen , 1893 .

[22]  Dima Grigoriev,et al.  Complexity of Null-and Positivstellensatz proofs , 2001, Ann. Pure Appl. Log..

[23]  László Lovász,et al.  Interactive proofs and the hardness of approximating cliques , 1996, JACM.

[24]  Arist Kojevnikov,et al.  Lower Bounds of Static Lovász-Schrijver Calculus Proofs for Tseitin Tautologies , 2006, ICALP.

[25]  Massimo Lauria,et al.  Tight Size-Degree Bounds for Sums-of-Squares Proofs , 2015, computational complexity.

[26]  Jan Krajícek,et al.  Lower bounds on Hilbert's Nullstellensatz and propositional proofs , 1994, Proceedings 35th Annual Symposium on Foundations of Computer Science.

[27]  Barnaby Martin,et al.  Relativization makes contradictions harder for Resolution , 2013, Ann. Pure Appl. Log..

[28]  Toniann Pitassi,et al.  Communication lower bounds via critical block sensitivity , 2013, STOC.

[29]  利久 亀井,et al.  California Institute of Technology , 1958, Nature.

[30]  Yurii Nesterov,et al.  Squared Functional Systems and Optimization Problems , 2000 .

[31]  David Steurer,et al.  Sum-of-squares proofs and the quest toward optimal algorithms , 2014, Electron. Colloquium Comput. Complex..

[32]  Dima Grigoriev,et al.  Complexity of Positivstellensatz proofs for the knapsack , 2002, computational complexity.

[33]  Yuan Zhou,et al.  Approximability and proof complexity , 2012, SODA.

[34]  Jan Krajícek.  Combinatorics of first order structures and propositional proof systems , 2004, Arch. Math. Log..

[35]  Grant Schoenebeck,et al.  Linear Level Lasserre Lower Bounds for Certain k-CSPs , 2008, 2008 49th Annual IEEE Symposium on Foundations of Computer Science.

[36]  Subhash Khot,et al.  On the power of unique 2-prover 1-round games , 2002, Proceedings 17th IEEE Annual Conference on Computational Complexity.

[37]  Ryan O'Donnell,et al.  SOS Is Not Obviously Automatizable, Even Approximately , 2016, ITCS.