Lower Bounds on the Size of Semidefinite Programming Relaxations

@article{Lee2015LowerBO,
  title={Lower Bounds on the Size of Semidefinite Programming Relaxations},
  author={James R. Lee and Prasad Raghavendra and David Steurer},
  journal={Proceedings of the forty-seventh annual ACM symposium on Theory of Computing},
  year={2015}
}
We introduce a method for proving lower bounds on the efficacy of semidefinite programming (SDP) relaxations for combinatorial problems. In particular, we show that the cut, TSP, and stable set polytopes on n-vertex graphs are not the linear image of the feasible region of any SDP (i.e., any spectrahedron) of dimension less than 2nδ, for some constant δ > 0. This result yields the first super-polynomial lower bounds on the semidefinite extension complexity of any explicit family of polytopes… 
View on ACM
arxiv.org
168 Citations
Highly Influential Citations
6
Background Citations
40
Methods Citations
18
Results Citations
6

168 Citations

Citation Type

Citation Type

Maximum semidefinite and linear extension complexity of families of polytopes
TLDR
This paper relates the maximum semidefinite and linear extension complexity of a family of polytopes to the cardinality of this family and the minimum pairwise Hausdorff distance of its members and shows that the linear extension complex of every d-dimensional 0/1-polytope is bounded from above by O(2dd).
Exploring the bounds on the positive semidefinite rank
TLDR
Some existing bounds on the PSD-rank are explored and proved to be upper bounded by the polynomials of the regular rank of the matrix, which is equal to the dimension of the polytope (up to an additive constant).
Sum-of-squares hierarchy lower bounds for symmetric formulations
TLDR
The main technical theorem allows the study of the positive semidefiniteness to be reduced to the analysis of “well-behaved” univariate polynomial inequalities and gives a short elementary proof of Grigoriev/Laurent lower bound for finding the integer cut polytope of the complete graph.
Sum-of-Squares Hierarchy Lower Bounds for Symmetric Formulations
TLDR
This work introduces a method for proving Sum-of-Squares SoS/ Lasserre hierarchy lower bounds when the initial problem formulation exhibits a high degree of symmetry, and gives a short elementary proof of Grigoriev/Laurent lower bound for finding the integer cut polytope of the complete graph.
Extended Formulation Lower Bounds for Combinatorial Optimization
TLDR
This dissertation will show that any linearprogramming relaxation for refuting random instances of constraint satisfaction problems requires super-polynomial size, and shows that any symmetric semidefinite programming relaxation for the matching problemin general graphs requires exponential size.
The Matching Problem Has No Fully Polynomial Size Linear Programming Relaxation Schemes
TLDR
It turns out that the high extension complexity for the matching problem stems from the same source of hardness as in the case of the correlation polytope: a direct sum structure.
Size-degree trade-offs for sums-of-squares and positivstellensatz proofs
We show that if a system of degree-k polynomial constraints on n Boolean variables has a Sums-of-Squares (SOS) proof of unsatisfiability with at most s many monomials, then it also has one whose…
Strengths and Limitations of Linear Programming Relaxations
TLDR
This thesis addresses both the power and the limitations of linear programming relaxations, as well as the connection between the shortcomings of these relaxations and the inapproximability of the underlying problem, and obtains LP relaxations of quasi-polynomial size that are at least as strong as that given by the knapsack cover inequalities.
Lower Bounds for Approximating the Matching Polytope
We prove that any extended formulation that approximates the matching polytope on nvertex graphs up to a factor of (1 + ε) for any 2 n ≤ ε ≤ 1 must have at least ( n α/ε ) defining inequalities where…
No Small Linear Program Approximates Vertex Cover within a Factor 2 -- e
TLDR
This work proves the following unconditional result about linear programming (LP) relaxations of the vertex cover problem: every LP relaxation that approximates vertex cover within a factor of 2 - ε has super-polynomially many inequalities.
...
1
2
3
4
5
...

References

SHOWING 1-10 OF 66 REFERENCES
Towards Sharp Inapproximability For Any 2-CSP
  • Per Austrin
  • Computer Science
    48th Annual IEEE Symposium on Foundations of Computer Science (FOCS'07)
  • 2007
TLDR
It is shown how to reduce the search for a good inapproximability result to a certain numeric minimization problem, and conjecture that the restricted type required for the hardness result is in fact no restriction, which would imply that these upper and lower bounds match exactly.
On the Power of Symmetric LP and SDP Relaxations
TLDR
For k <; n/4, it is shown that k-rounds of sum-of squares / Lasserre relaxations of size k(kn) achieve best-possible k approximation guarantees for Max CSPs among all symmetric SDP relaxation of size at most (kn).
Optimal algorithms and inapproximability results for every CSP?
TLDR
A generic conversion from SDP integrality gaps to UGC hardness results for every CSP is shown, which achieves at least as good an approximation ratio as the best known algorithms for several problems like MaxCut, Max2Sat, MaxDiCut and Unique Games.
Approximability and proof complexity
TLDR
This work considers bounded-degree "Sum of Squares" (SOS) proofs, a powerful algebraic proof system introduced in 1999 by Grigoriev and Vorobjov, and shows that this proof is automatizable using semidefinite programming (SDP), meaning that any n-variable degree-d proof can be found in time nO(d).
The matching polytope has exponential extension complexity
TLDR
By a known reduction this also improves the lower bound on the extension complexity for the TSP polytope from 2Ω(√n) to 2 Ω(n).
Polylogarithmic inapproximability
TLDR
It is shown that for every fixed ε>0, the GROUP-STEINER-TREE problem admits no efficient log2-ε k approximation, where k denotes the number of groups (or, alternatively, the input size), unless NP has quasi polynomial Las-Vegas algorithms.
On the existence of 0/1 polytopes with high semidefinite extension complexity
TLDR
It is shown that there is a 0/1 polytope such that any spectrahedron projecting to it must be the intersection of a semidefinite cone of dimension $$2^{\varOmega (n)}$$2Ω(n) and an affine space.
Hypercontractivity, sum-of-squares proofs, and their applications
TLDR
Reductions between computing the 2->4 norm and computing the injective tensor norm of a tensor, a problem with connections to quantum information theory and the study of Khot's Unique Games Conjecture are shown.
Equivariant Semidefinite Lifts and Sum-of-Squares Hierarchies
TLDR
A representation-theoretic framework is presented to study equivariant PSD lifts of a certain class of symmetric polytopes known as orbitopes which respect the symmetries of the polytope.
Approximation Limits of Linear Programs (Beyond Hierarchies)
TLDR
A quantitative improvement of Razborov's rectangle corruption lemma for the high error regime gives strong lower bounds on the nonnegative rank of certain perturbations of the unique disjoint ness matrix that proves that quadratic approximations for CLIQUE require linear programs of exponential size.
...
1
2
3
4
5
...