TBA

TBA

Through the process of tropicalization one obtains from an algebraic variety $X$ a combinatorial object called the tropicalization of $X$, $trop(X)$, that retains a lot of information about the original variety. Following the work of J. Giansiracusa and N. Giansiracusa, one can endow $trop(X)$ with more structure, to obtain a tropical scheme. Loosely speaking, we consider more equations than the ones needed to determine the tropical variety. We are interested what information about the original variety $X$ is preserved by the tropical scheme $X_\mathbb{T}$ (but possibly not by the tropical variety). In particular, we study the relation between the Picard group of $X$ and $X_\mathbb{T}$. We solve the problem in the case when $X$ is a toric variety.

It is due to Blekherman that we now know there are significantly more nonnegative forms than sums of squares for any fixed degree \(2d\) and large number of variables \(n\). What if we ask the same comparison for polynomials with a prescribed Newton polytope? I will present some general ideas from (modern) convex geometry that handles the case of multihomogeneous polynomials. These results refine the estimates of Blekherman and happens to have some implications in quantum information theory. I will also try to indicate missing ingredient(s) for handling general Newton polytopes, which might be of interest to the lovers of toric geometry and/or Lie algebras.

I will discuss some methods/techniques to prove "small ball probabilities" and I will review some connections with other problems in Asymptotic Geometric Analysis. The emphasis will be on how high-dimensional geometry and convexity is crucial to understand the concentration behavior in the "small ball regime".

An affine semigroup \(S\) is a subset of \(\mathbb Z_{\ge 0}^k\) that is closed under vector addition, and a factorization of \(a \in S\) is an expression of \(a\) as a sum of generators of \(S\). The delta set of \(a\) is a set of positive integers determined by the 'missing factorization lengths' of \(a\), and the delta set of \(S\) is the union of the delta sets of its elements. Although the delta set of any affine semigroup is finite, its definition as an infinite union makes explicit computation difficult. In this talk, we explore algebraic and geometric properties of the delta set, and survey the history of its computation for affine semigroups. The results presented here span the last 20 years, ranging from a first algorithm for a small class of semigroups that is impractical for even basic examples, to recent joint work with Garcia-Sanchez and Webb expressing the delta set of any affine semigroup in terms of Groebner bases, and include results from numerous undergraduate research projects.

To compute certificates of nonnegativity, an approach based on sums of squares (SOS) decompositions has been popularized by Lasserre and Parillo.
In the first part of this talk, we consider the problem of finding SOS decompositions of nonnegative univariate polynomials. It is well-known that every non-negative univariate real polynomial can be written as the sum of two polynomial squares with real coefficients. When one allows a weighted sum of finitely many squares instead of a sum of two squares, then one can choose all coefficients in the representation to lie in the field generated by the coefficients of the polynomial. We describe, analyze and compare two algorithms computing such a weighted sums of squares decomposition for univariate polynomials with rational coefficients. - The first algorithm, due to Schweighofer, relies on real root isolation, quadratic approximations of positive polynomials and square-free decomposition but its complexity was not analyzed. We provide bit complexity estimates, both on runtime and output size of this algorithm. They are exponential in the degree of the input univariate polynomial and linear in the maximum bitsize of its complexity. This analysis is obtained using quantifier elimination and root isolation bounds. - The second algorithm, due to Chevillard, Harrison, Joldes and Lauter, relies on complex root isolation and square-free decomposition and has been introduced for certifying positiveness of polynomials in the context of computer arithmetics. Again, its complexity was not analyzed. We provide bit complexity estimates, both on runtime and output size of this algorithm, which are polynomial in the degree of the input polynomial and linear in the maximum bitsize of its complexity. This analysis is obtained using Vieta's formula and root isolation bounds. While the second algorithm is, as expected from the complexity result, more efficient on most of examples, we exhibit families of non-negative polynomials for which the first algorithm is better.
In the second part of this talk, we consider the problem of finding exact sums of squares (SOS) decompositions for certain classes of nonnegative multivariate polynomials, while relying on semidefinite programming (SDP) solvers. - We start by providing a hybrid symbolic-numeric algorithm computing exact rational SOS decompositions for polynomials lying in the interior of the SOS cone. This algorithm computes an approximate SOS decomposition for a perturbation of the input polynomial with an arbitrary-precision SDP solver. An exact SOS decomposition is obtained thanks to the perturbation terms. We prove that bit complexity estimates on output size and runtime are both polynomial in the degree of the input polynomial and simply exponential in the number of variables. This analysis is based on quantifier elimination as well as bounds on the cost of the ellipsoid method and Cholesky's decomposition. - Next, we apply this algorithm to compute exact Polya and Putinar's representations respectively for positive definite forms and polynomials positive over basic compact semialgebraic sets. We compare the implementation of our algorithms with existing methods based on CAD or critical points.
This is joint work with Mohab Safey El Din and Markus Schweighofer.

Buildings form a fundamental geometric concept which generalize the Coxeter complex of a Coxeter group. The theory was mostly developed by Tits (and Borel) in the 1960s and 1970s. While its origins are in Lie theory one particularly striking application was its contribution to the classification of the finite simple groups (completed around 1980). In this talk (of historical survey type) I will define the three notions in the title and look into one family of examples. In the end I will briefly try to sketch the impact on group theory in general.

Constructing efficient algorithms for combinatorial optimization problems is a task that requires a lot of knowledge, experience and creativity. The goal of this talk is to present the automatizable hierarchies of algorithms (Lasserre/Sum-of-Squares (SoS), Sherali Adams (SA)) and the recent development in this field, the Sum-of-Nonnegative-Circuit (SoNC) algorithm.
We start with a motivation and a gentle introduction to the hierarchies of relaxations for 0/1 optimization problems. Next we discuss the strength and weaknesses of this approach and present a novel method in this field, the SoNC hierarchy.
Finally, in this talk we show that key results for SOS/SA hierarchy remain valid for SoNC hierarchy. More precisely we show that for a polynomial optimization problem over the $n$-variate boolean hypercube with constraints of degree $d$, the SoNC hierarchy at level $n+d$ converges to the convex hull of integral solutions.

Fuglede's conjecture (1974) states that a bounded measurable subset in \(\mathbb{R}^d\) accepts an orthogonal basis of exponential functions (i.e. it is spectral) if and only if it tiles the space with a discrete set of translations. This conjecture turned out to be false by Tao's counterexample in 2003. Using Tao's ideas, counterexamples in finite Abelian groups such as \(\mathbb{Z}_N^d\) can be lifted to counterexamples in \(\mathbb{R}^d\), thus shifting the interest on this conjecture to this setting in recent years. This has been successful for \(d\geq3\), but the conjecture is still open for \(d=1,2\).
Some recent results in the cyclic group setting will be presented in this talk, which are connected to the work of Coven-Meyerowitz and Laba on tiling subsets of \(\mathbb{Z}\), as well as the structure of vanishing sums of roots of unity. This is joint work with Mihalis Kolountzakis & work in progress.

In this talk we explain how every even, zonal measure on the Euclidean unit sphere gives rise to a sharp Sobolev inequality for functions of bounded variation which directly implies the classical Euclidean Sobolev inequality. The strongest member of this large family of inequalities is shown to be the only affine invariant one among them – the affine Zhang-Sobolev inequality. We discuss in some detail the geometry behind these analytic inequalities and also relate our new Sobolev inequalities to the sharp Gromov-Gagliardo-Nirenberg Sobolev inequality for general norms and discuss further improvements of special cases.

Moment-sum-of-squares hierarchies of semidefinite programs can be used to approximate the volume of a given compact basic semialgebraic set K. The idea consists of approximating from above the indicator function of K with a sequence of polynomials of increasing degree d, so that the integrals of these polynomials generate a convergence sequence of upper bounds on the volume of K. In this talk we show how we could derive an asymptotic rate of convergence for this approximation scheme. Joint work with Milan Korda, see arXiv:1609.02762.

In this talk we present the SoS hierarchy in the context of 0/1 optimization along with some of our recent integrality gap/lower bound results. We introduce the relevant concepts and give a brief derivation of the hierarchy. As for our results, we characterize the possible gap instances on the round n-1 of the hierarchy. Then we show an unbounded integrality gap after nonconstant number of rounds for a problem that admits a polynomial time algorithm. Finally we present a simplification of the positive semidefiniteness constraints when a high degree of symmetry can be assumed, and some applications to integrality gaps and lower Bounds.

We present structural results on sparse nonnegative integer solutions to underdetermined systems of linear equations.
Our proofs are algebraic and number theoretic in nature and make use of Siegel's Lemma and classical tools from the Geometry of Numbers.
These results have some interesting consequences in discrete optimisation.

In this talk we will present algorithms to check smoothness for the class of spherical varieties which contains those of toric varieties, flag varieties and symmetric varieties, and which forms a remarkable class of algebraic varieties with an action by an algebraic group having an open dense orbit. In the toric case there is a well-known simple combinatorial smoothness criterion whereas in the spherical case Brion, Camus and Gagliardi have shown smoothness criteria for spherical varieties which either are rather involved or rely on certain classification results. We suggest a purely combinatorial smoothness criterion by introducing a rational invariant which only depends on the combinatorics of the spherical variety. We have a conjectural inequality this invariant should satisfy where the equality case would imply that the spherical variety has a toric torus action. Our conjecture would also imply the generalized Mukai conjecture for spherical varieties. We complete our talk by summarizing in which cases the above mentioned approach is known to be true. This is joint work with Giuliano Gagliardi.

Given \(k\in \mathbb{R}_{\ge 0}\) and a vector \(v\) of Plücker coordinates of a point in the real Grassmannian, is the vector obtained by taking the \(k\)th power of each entry of \(v\) again a vector of Plücker coordinates? We will show that for \(k\neq 1\), this is true if and only if the corresponding matroid is regular. Similar results hold over other fields. We will also describe the subvariety of the Grassmannian that consists of all the points that define a regular matroid.
An arithmetic matroid is a matroid equipped with a function that assigns a multiplicity (a positive integer) to each subset of the ground set. We will discuss a discrete version of the problem described above that deals with the representability of the arithmetic matroid obtained by taking the \(k\)th power of the multiplicity function. This setting is motivated by some combinatorial questions on CW complexes.

We live in the age of data, huge amount of Data its been generated and acquired everyday in different forms and flavors. One would like to have a technique to store and process optimally this data; this is possible since the relevant information in almost all data found in typical applications is sparse due the high correlation of its elements. The challenge will lie in the search of appropriate dictionary that can represent optimally this information. Along the history very talented scientist have proposed different signal transforms based on certain representation dictionaries, starting with the Fourier Transform and Short Time Fourier Transform; in 1980s the wavelet transform was proposed representing a breakthrough in one dimensional signal representation, with the ability to not just optimally represent the data but also capture certain singularities, regularities and other features; the flaw of the wavelets transforms is its lack of directional sensitivity which does not aloud it to represent optimally multidimensional singularities like curves that are predominant in general multidimensional signals (for instance images and videos). The Shearlet transform is a product of almost 10 years of research on sparsifying transform with directional sensitivity that attains the optimal best N-term approximation error, it was proposed by G. Kutyniok, D. Labate and K Guo in 2015 and since then it's been known to be the best of its type.
In this talk I will explain the upsides and downsides of the already mentioned transforms, why the shearlet transform is great solution for anisotropic representation of multidimensional signals, how can we implement wavelet and shearlet transform in Julia, and I will show you some applications and demos of the Shearlet Transform Library that I implemented with Professor Gitta Kutyniok based on the matlab version developed by her group two years ago; I will also explain why Julia is the best option to implement algorithms in Signal and Image Processing and show some benchmarks that show its great performance.

A family \(\mathcal{F}\) of sets is in convex position if none of its members is contained in the convex hull of the union of the others. The members of \(\mathcal{F}\) are ovals (compact convex sets) in the plane that have a certain property. An Erdos-Szekeres type theorem concerns the existence, for any integer \(n\geq 3\), of a smallest positive integer \(N(n)\) such that if \(|\mathcal{F}|≥N(n)\) then there are \(n\) ovals of \(\mathcal{F}\) in convex position.
In this talk, I survey some known theorems, introduce a new one based upon work with Gabor Fejes Toth and examine the relation between \(N(n)\) and Ramsey numbers of the type \(R_k(n_1,n_2)\).

This talk will focus on open problems in Geometric Tomography, which aims to retrieve information about a geometric object (such as a convex body, star body, finite set, etc.) from data concerning its intersections with planes or lines and/or projections (i.e., shadows) on planes or lines. The problems, which span nearly a hundred years of mathematics, are diverse. Many have a common thread, however, since they are linked to various integral transforms: the X-ray transform, divergent beam transform, circular Radon transform, cosine transform, or spherical Radon transform.
The talk will be illustrated by plenty of pictures.

Numerical computations by Cohn, Kumar, and Schürmann in energy minimizing periodic configurations of density \(\rho\) and \(1/\rho\) revealed an impressive kind of symmetry, the so–called formal duality. Formal dual subsets of \(\mathbb{R}^n\) satisfy a generalized version of Poisson summation formula, and it is conjectured that the only periodic subsets of \(\mathbb{R}\) of density \(1\) possessing a formal dual subset, are \(\mathbb{Z}\) and \(2\mathbb{Z}\cup(2\mathbb{Z}+\frac12)\).
In this talk, we will present recent progress on this conjecture by the speaker, utilizing several tools from different areas of mathematics, such as (a) the field descent method by Bernhard Schmidt, used chiefly on the circulant Hadamard matrix conjecture, (b) the 'polynomial method' by Coven & Meyerowitz used in the characterization of sets tiling \(\mathbb{Z}\) or \(\mathbb{Z}_N\), and (c) basic arithmetic of cyclotomic fields.

We introduce mplrs, a new parallel vertex/facet enumeration program based on the reverse-search code lrs. The implementation uses MPI and can be used on single machines, clusters and supercomputers. We describe the budgeted parallel tree search implemented in mplrs and compare performance with other sequential and parallel programs for vertex/facet enumeration. In some instances, mplrs achieves almost linear scaling with more than 1000 cores. The approach used in mplrs can be easily applied to various other problems. This is joint work with David Avis.

A lattice polytope is a convex polytope each of whose vertices has integer coordinates. It is known that a lattice simplex of dimension \(d\) corresponds to a finite abelian subgroup of \((\mathcal{R}/\mathcal{Z})^{d+1}\). Conversely, given a finite abelian subgroup of \((\mathcal{R}/\mathcal{Z})^{d+1}\) such that the sum of all entries of each element is an integer, we can obtain a lattice simplex of dimension \(d\). In this talk, we discuss a characterization of Gorenstein simplices in terms of the associated finite abelian groups. Gorenstein polytopes are of interest in combinatorial commutative algebra, mirror symmetry and tropical geometry. In particular, we present complete characterizations of Gorenstein simplices whose normalized volume equal \(p\), \(p^2\) and \(pq\), where \(p,q\) are prime integers.

Although hyperplane arrangement complements are rationally formal, we note that they have non-minimal rational (CDGA) models which are topologically and combinatorially significant. We construct a family of CDGAs which interpolates between the Orlik-Solomon algebra and the cohomology algebras of arrangement compactifications. Our construction is combinatorial and extends to all matroids, regardless of their (complex) realizability.
This is joint work with Christin Bibby and Graham Denham.

I present the recent work 'Slicing and dicing polytopes' (arXiv:1608.05372) by Patrik Norén on cellular resolutions. I introduce diced and sharp polytopes which are necessary for the construction and give a brief introduction to cellular resolutions. The main result, introducing a new class of polyhedral subdivisions which support cellular resolutions, is illustrated along some examples.

By studying the volume of a generalized difference body, this talk presents the first nontrivial lower bound for the lattice covering density by \(n\)-dimensional simplices.

tba

A classification of symmetric tensor valuations on lattice polytopes in \(\mathbb{R}^n\) that intertwine the special linear group over the integers is established for the cases in which the rank of the tensor is at most \(n\). The scalar-valued case was classified by Betke and Kneser where it was shown that the only such valuations are the coefficients of the Ehrhart polynomial. Extending this result, the coefficients of the discrete moment tensor form a basis for the symmetric tensor valuations.

The Hardy constant of a simply connected domain \(\Omega\subset\mathbb{R}^2\) is the best constant for the inequality \[ \int_\Omega |\nabla u|^2 dx \geq c\int_\Omega \frac{u^2}{\operatorname{dist}(x,\partial\Omega)^2} dx, \quad u\in C_c^\infty(\Omega). \] After the work of Ancona where the universal lower bound 1/16 was obtained, there has been a substantial interest on computing or estimating the Hardy constant of planar domains. In this talk we determine the Hardy constant of an arbitrary quadrilateral as well as of some other planar domains (joint work with Achilles Tertikas).

The main aim of compressed sensing is to recover a low-dimensional object (e.g. a sparse vector, a low-rank matrix, ...) from few linear measurements. Many algorithms for this task (e.g. Basis Pursuit, nuclear norm-minimization) consists of minimizing a structure-promoting function \(f\) over the set of possible solutions.

An elegant approach to finding recovery guarantees of such programs relies on high-dimensional geometry of convex cones. The techniques were developed relatively recently, but have quickly gained popularity. In the first half of this talk, this approach will be presented in relative high detail.

In the second half of the talk, we will discuss the connection between geometry and stability of compressed sensing algorithms. In particular, we will present a new geometric criterion on \(A\) implying stability both with respect to model inexactness of the signal as well as to additive noise in the measurements.

The spectral theory of tensors aims at generalizing the concept of eigenvalues and eigenvectors of matrices to higher dimensional arrays. In this talk I will focus on the definition of E-eigenvalues and Z-eigenvalues given by Qi. After giving the definition itself I will explain various applications, give the higher dimensional equivalent of the characteristic polynomial and talk about the number of eigenvalues of both real and complex tensors. The last part of the talk will be about an algorithm to compute eigenvalues of tensors using homotopy methods and the complexity of this algorithm.

Given a subset S of R^n, let c(S,k) be the smallest number t such that whenever finitely many convex sets have exactly k common points in S, there exist at most t of these sets that already have exactly k common points in S. For S = Z^n, this number was introduced by Aliev et al. [2014] who gave an explicit bound showing that c(Z^n,k) = O(k) holds for every fixed n. Recently, Chestnut et al. [2015] improved this to c(Z^n,k) = O(k (log log k)(log k)^{-1/3} ) and provided the lower bound c(Z^n,k) = Omega(k^{(n-1)/(n+1)}). We provide a combinatorial description of c(S,k) in terms of polytopes with vertices in S and use it to improve the previously known bounds as follows: We strengthen the bound of Aliev et al. [2014] by a constant factor and extend it to general discrete sets S. We close the gap for Z^n by showing that c(Z^n,k) = Theta(k^{(n-1)/(n+1)}) holds for every fixed n. Finally, we determine the exact values of c(Z^n,k) for all k <= 4.

In this talk, we study the problem of estimating a structured signal \(x_0\in\mathbb{R}\) linear Gaussian observations. Supposing that \(x_0\) belongs to a certain convex subset \(K\subset\mathbb{R}\) we will see that an accurate recovery is possible as long as the number of observations exceeds the effective dimension of \(K\), which is a common measure for the complexity of signal classes. Interestingly, it will turn out that the (possibly unknown) non-linearity of our model affects the error rate only by a multiplicative constant. This achievement is based on recent works by Yaniv Plan and Roman Vershynin, who have suggested to treat the non-linearity rather as noise which perturbs a linear measurement process. Using the concept of restricted strong convexity, we show that their results for the generalized Lasso can be extended to a fairly large class of convex loss functions. This is especially appealing for practical applications, since in many real-world scenarios, adapted loss functions empirically perform better than the classical square loss. To this end, the presented results provide a unified and general framework for signal reconstruction in high dimensions, covering various challenges from the fields of compressed sensing, signal processing, and statistical learning.

This talk is a summary of the very recent paper by Matthew Baker of the same title (http://arxiv.org/pdf/1601.01204.pdf). Various flavours of ''matroids with additional structure'', such as valuated and oriented matroids have been around for quite a while, with often very little connection between the different matroid species. Matroids over hyperfields are a generalization of all those concepts - thus allowing a unified treatment. In this talk I will first give a short introduction to hyperfields. I will then define what matroids over hyperfields are, give examples and discuss generalizations of familiar matroid operations such as duality and minors.

Wished for by Weil and defined by Grothendieck, Étale cohomology is to varieties over fields of finite characteristic what usual cohomology is to complex or real algebraic varieties when they are equipped with the analytic, respectively Euclidean topology. This very short introduction will motivate and try to explain the basic ingredients of Étale cohomology. Throughout I will make analogies to the complex and real situations and stick to simple examples in hopes to provide some intuition into this deep theory.

The Minkowski measure of symmetry s(K) of a convex body K, is the smallest positive dilatation of K containing a translate of -K. In this talk we will explain some of its basic properties in detail.

Afterwards, we will show how s(.) can be used to strengthen, smoothen, and join different geometric inequalities, as well as its connections to other concepts such as diametrical compleness, Jung's inequality, or Banach-Mazur distance.

The Turaev-Viro invariants are a powerful family of topological invariants for distinguishing between different 3-manifolds. They are invaluable for mathematical software, but current algorithms to compute them require exponential time. I will discuss this family of invariants, and present an explicit fixed-parameter tractable algorithm for arbitrary r which is practical -- and indeed preferable -- to the prior state of the art for real computation.
This is joint work with Ben Burton and Clément Maria