arXiv Science⌕ Search

arXiv subjects

Yushan Qu

Publications and source records attributed to Yushan Qu.

6 recordsLinked to original sources

Arithmetic Universality and Integer Decomposition in Disjunctions of Three Staircase Simplices

The staircase simplices $\{x\in[0,1]^d:x_{σ(1)}\le\cdots\le x_{σ(d)}\}$, $σ\in S_d\,$, are the $d!$ unimodular cells of the staircase triangulation of the unit cube. The disjunctive hull (Cayley polytope) of a selection $P_0, \ldots, P_n$ of them has the integer decomposition property (IDP) if, for all integers $λ_i\ge0$, every lattice point of $\sum_iλ_i P_i$ is a sum of vertices, $λ_i$ of them from $P_i$. For two simplices, IDP always holds. For three, every nontrivial finite abelian group $G$ arises, in the optimal dimension $Θ(\log|G|)$, as the residue group of a simplex facet whose $|G|-1$ nonzero box points are Hilbert basis elements of the whole cone; $k$ times such a point is a sum of vertices exactly when its order in $G$ divides $k$. Isolating this arithmetic is costly: if a simplex face has a single Hilbert basis element $h$ that is not a vertex, then $kh$ is a sum of vertices exactly when some $q$ divides $k$, and $d\ge3q-2$, attained for every $q\ge3$; we also find the least dimension for each $G$ with the fewest such elements. Without isolation, every nonzero submonoid of $\mathbb{Z}_{\ge0}$ omitting $1$ is $\{k\ge0:kh$ is a sum of vertices$\}$ for a Hilbert basis element $h$ of degree three. For every $0/1$ polytope with $N$ vertices in $\mathbb{R}^r$, three staircase simplices in dimension $O(N(r+1))$ have a face that has IDP exactly when the polytope does. Deciding whether a lattice point $h$ is a sum of vertices is strongly NP-complete, even given vertex expressions of $2h$ and $3h$, and recognizing IDP of a designated facet is coNP-complete; for the whole hull, IDP can first fail in degree linear in $d$, and it holds whenever $d\le5$. For a family of triples defined by words over three letters, we determine the entire Hilbert basis and characterize IDP by two forbidden subsequences.

math.CO↗

On the facets of the Cayley polytope of Stanley's triangulation

Disjunctions over the cells of a triangulation arise in formulations of piecewise-linear functions. For a poset $P$ on $[d]$, we study the convex hull $D$ of the simplices of Stanley's triangulation of the order polytope of $P$, each lifted by its own indicator vector. Via the Cayley trick, for any disjunction of full-dimensional polytopes, the facets of the convex hull other than those described by the bounds on the indicator variables correspond to the facet normals of the Minkowski sum of the polytopes, and each is described by the full optimal big-M lifting of the supporting inequality with that normal of any one polytope. For Stanley's triangulation, with $M(P)$ the Minkowski sum of its simplices, lifting only the facet-describing inequalities of the simplices need not suffice, unlike some other cases; we show that it suffices if and only if $M(P)$ is alcoved, if and only if deleting one element of $P$ leaves a chain. Otherwise, the coefficients of the additional facet-describing inequalities are governed by the least number $\ell(P)$ of elements whose deletion leaves a chain: for fixed $\ell(P)$, the largest entry of a primitive facet normal grows like $d^{\ell(P)-1}$, and it can be exponential in $d$ if $\ell(P)$ grows with $d$. For $d \geq 3$ and $\ell(P)\leq2$ the sharp bound is $d-2$.

math.CO↗

Facet normals of Minkowski sums of staircase simplices

Staircase simplices are the simplices of the standard (Freudenthal--Kuhn) triangulation of $[0,1]^d$ and of Stanley's triangulations of order polytopes. We study the facet normals of their Minkowski sums and how the number of summands constrains the size of primitive integer normals. We characterize the normals attainable with a prescribed number of summands in terms of $d-1$ linearly independent zero-one vectors orthogonal to the normal, whose supports have an intersection graph that can be colored with one color per summand. This gives a combinatorial characterization of the least number of summands needed to realize a normal. For two summands, all primitive normals have entries in $\{-1,0,1\}$, and we determine which of them occur. For three summands we bound every entry in absolute value by $3^{(d-2)/3}$ and construct normals with entries of order $θ^d$, $θ$ the plastic number; for four summands a second construction gives entries of order $ψ^d$, $ψ$ the supergolden ratio. Since $ψ>3^{1/3}$, the largest coefficients for three and four summands are exponentially separated. For vectors with at most two distinct nonzero entry values, we determine the least number of summands exactly, using a matching-rank formula proved by exterior algebra and representations of symmetric groups. With no restriction on the number of summands, the maximum coefficient equals the classical maximum for facets of $0/1$-polytopes in the same dimension.

math.CO↗

More on Disjunctive Convex Hulls of Generalized Cross Polytopes

We continue the study of the convex hull $D\subset\mathbb{R}^{d+n}$ of a disjunction of $n+1$ polytopes in $\mathbb{R}^d$, managed by $n$ binary indicator variables. Previously, $D$ had been characterized for arbitrary polytopes when $d\in\{1,2\}$, for polytopes sharing a common constraint matrix (under a technical condition, including hyper-rectangles), and for two generalized cross polytopes ($n=1$). Here we give a complete description of $D$ for arbitrary $n$ and $d$ when all $n+1$ polytopes are generalized cross polytopes. Every vertical facet is described by a normalized support inequality, and such an inequality is facet describing if and only if its equality graph, a bipartite graph on the polytopes and the coordinates, is connected; all vertical facets are generated explicitly from feasible spanning trees. We characterize when such a facet is one of the previously known pairwise inequalities, and we show that pairwise inequalities suffice when $d\leq3$, but not in general. We establish a bijection between the vertical facets and the maximal cells of a regular subdivision of $Δ_n\timesΔ_{d-1}$\,, which yields the sharp bound $n+1+2^d\binom{n+d-1}{n}$ on the number of facets of $D$, attained for generic data. We also give a polynomial-time separation algorithm that returns a violated facet, which is maximally violated among the normalized support inequalities. Computational experiments show that the facets that are not pairwise inequalities close a large part of the gap left by the pairwise inequalities.

math.OC↗

On disjunction convex hulls for generalized cross polytopes

We continue the study of the natural polytope $\mathcal{D}$ in $\mathbb{R}^{n+d}$ associated with the disjunction of a set of $n+1$ polytopes in $\mathbb{R}^d$, managed by $n$ binary variables. Already $\mathcal{D}$ had been characterized for arbitrary $n\geq 1$ and (i) $d\in\{1,2\}$, and (ii) for a broad generalization of hyper-rectangles. In both cases, the complete characterization employs full optimal big-M lifting. Here, we give a complete description of $\mathcal{D}$ for the case of $n=1$ and arbitrary $d$, when the (two) polytopes are arbitrary generalized cross polytopes. Furthermore, we characterize when our complete description employs only optimal big-M lifting. For $n>1$, we generalize the family of facet-describing inequalities used for $n=1$. Finally, we carry out some computational experiments demonstrating the value of our theoretical results.

math.OC↗

On disjunction convex hulls by lifting

We study the natural extended-variable formulation for the disjunction of $n+1$ polytopes in $\mathbb{R}^d$. We demonstrate that the convex hull $D$ in the natural extended-variable space $\mathbb{R}^{d+n}$ is given by full optimal big-M lifting (i) when $d\leq 2$ (and that it is not generally true for $d\geq 3$), and also (ii) under some technical conditions, when the polytopes have a common facet-describing constraint matrix, for arbitrary $d\geq 1$ and $n\geq 1$. We give a broad family of examples with $d\geq 3$ and $n=1$, where the convex hull is not described after employing all full optimal big-M lifting inequalities, but it is described after one round of MIR inequalities. Additionally, we give some general results on the polyhedral structure of $D$, and we demonstrate that all facets of $D$ can be enumerated in polynomial time when $d$ is fixed.

math.OC↗