arXiv subjects

Kai Kellner

Publications and source records attributed to Kai Kellner.

6 recordsLinked to original sources

arXiv2019

Irreducible infeasible subsystems of semidefinite systems

Farkas' lemma for semidefinite programming characterizes semidefinite feasibility of linear matrix pencils in terms of an alternative spectrahedron. In the well-studied special case of linear programming, a theorem by Gleeson and Ryan states that the index sets of irreducible infeasible subsystems are exactly the supports of the vertices of the corresponding alternative polyhedron. We show that one direction of this theorem can be generalized to the nonlinear situation of extreme points of general spectrahedra. The reverse direction, however, is not true in general, which we show by means of counterexamples. On the positive side, an irreducible infeasible block subsystem is obtained whenever the extreme point has minimal block support. Motivated by results from sparse recovery, we provide a criterion for the uniqueness of solutions of semidefinite block systems.

math.OC↗

arXiv2016

A Semidefinite Hierarchy for Disjointly Constrained Multilinear Programming

Disjointly constrained multilinear programming concerns the problem of maximizing a multilinear function on the product of finitely many disjoint polyhedra. While maximizing a linear function on a polytope (linear programming) is known to be solvable in polynomial time, even bilinear programming is NP-hard. Based on a reformulation of the problem in terms of sum-of-squares polynomials, we study a hierarchy of semidefinite relaxations to the problem. It follows from the general theory that the sequence of optimal values converges asymptotically to the optimal value of the multilinear program. We show that the semidefinite hierarchy converges generically in finitely many steps to the optimal value of the multilinear problem. We outline two applications of the main result. For nondegenerate bimatrix games, a Nash equilibrium can be computed by the sum of squares approach in finitely many steps. Under an additional geometric condition, the NP-complete containment problem for projections of $\mathcal{H}$-polytopes can be decided in finitely many steps.

math.OC↗

arXiv2016

Sum of Squares Certificates for Containment of $\mathcal{H}$-polytopes in $\mathcal{V}$-polytopes

Given an $\mathcal{H}$-polytope $P$ and a $\mathcal{V}$-polytope $Q$, the decision problem whether $P$ is contained in $Q$ is co-NP-complete. This hardness remains if $P$ is restricted to be a standard cube and $Q$ is restricted to be the affine image of a cross polytope. While this hardness classification by Freund and Orlin dates back to 1985, for general dimension there seems to be only limited progress on that problem so far. Based on a formulation of the problem in terms of a bilinear feasibility problem, we study sum of squares certificates to decide the containment problem. These certificates can be computed by a semidefinite hierarchy. As a main result, we show that under mild and explicitly known preconditions the semidefinite hierarchy converges in finitely many steps. In particular, if $P$ is contained in a large $\mathcal{V}$-polytope $Q$ (in a well-defined sense), then containment is certified by the first step of the hierarchy.

math.CO↗

arXiv2015

Containment Problems for Projections of Polyhedra and Spectrahedra

Spectrahedra are affine sections of the cone of positive semidefinite matrices which form a rich class of convex bodies that properly contains that of polyhedra. While the class of polyhedra is closed under linear projections, the class of spectrahedra is not. In this paper we investigate the problem of deciding containment of projections of polyhedra and spectrahedra based on previous works on containment of spectrahedra. The main concern is to study these containment problems by formulating them as polynomial nonnegativity problems. This allows to state hierarchies of (sufficient) semidefinite conditions by applying (and proving) sophisticated Positivstellensätze. We also extend results on a solitary sufficient condition for containment of spectrahedra coming from the polyhedral situation as well as connections to the theory of (completely) positive linear maps.

math.OC↗

arXiv2015

A Semidefinite Hierarchy for Containment of Spectrahedra

A spectrahedron is the positivity region of a linear matrix pencil and thus the feasible set of a semidefinite program. We propose and study a hierarchy of sufficient semidefinite conditions to certify the containment of a spectrahedron in another one. This approach comes from applying a moment relaxation to a suitable polynomial optimization formulation. The hierarchical criterion is stronger than a solitary semidefinite criterion discussed earlier by Helton, Klep, and McCullough as well as by the authors. Moreover, several exactness results for the solitary criterion can be brought forward to the hierarchical approach. The hierarchy also applies to the (equivalent) question of checking whether a map between matrix (sub-)spaces is positive. In this context, the solitary criterion checks whether the map is completely positive, and thus our results provide a hierarchy between positivity and complete positivity.

math.OC↗

arXiv2013

Containment problems for polytopes and spectrahedra

We study the computational question whether a given polytope or spectrahedron $S_A$ (as given by the positive semidefiniteness region of a linear matrix pencil $A(x)$) is contained in another one $S_B$. First we classify the computational complexity, extending results on the polytope/polytope-case by Gritzmann and Klee to the polytope/spectrahedron-case. For various restricted containment problems, NP-hardness is shown. We then study in detail semidefinite conditions to certify containment, building upon work by Ben-Tal, Nemirovski and Helton, Klep, McCullough. In particular, we discuss variations of a sufficient semidefinite condition to certify containment of a spectrahedron in a spectrahedron. It is shown that these sufficient conditions even provide exact semidefinite characterizations for containment in several important cases, including containment of a spectrahedron in a polyhedron. Moreover, in the case of bounded $S_A$ the criteria will always succeed in certifying containment of some scaled spectrahedron $νS_A$ in $S_B$.

math.OC↗