Events

Title: Stochastic  Oracles and Where to Find Them

Abstract: Continuous optimization is a mature field, which has recently undergone major expansion and change. One of the key new directions is the development of methods that do not require exact information about the objective function. Nevertheless, the majority of these methods, from stochastic gradient descent to "zero-th order" methods use some kind of approximate first order information. We will overview different methods of obtaining this information, including simple stochastic gradient via sampling, robust gradient estimation in adversarial settings, traditional and randomized finite difference methods and more.

We will discuss what key properties of these inexact, stochastic first order oracles are useful for convergence analysis of optimization methods that use them. 

Title: Adventures in Enumeration

Abstract: We make the argument that by combining pure mathematical tools with computational insights and applications from a vast array of disciplines, combinatorics is the perfect area to see all the wonders of math on display. Applications discussed include the analysis of classical algorithms, restricted permutations, models predicting the shape of biomembranes, queuing theory, random walks, ratchet models for gene expression, maximum likelihood degree in algebraic statistics, transcendence of zeta values, sampling algorithms for perfect matchings in bipartite graphs, and parallel synthesis for DNA storage.

Title: Reconstructing Shredded Random Matrices

Abstract: The Graph Reconstruction Conjecture states that if we are given the set of vertex-deleted subgraphs of some graph, then there is a unique graph G that can be reconstructed from them. This conjecture has been open since the 60s, and has only been solved for certain classes of graphs with not much progress towards the general case. We instead study adjacent reconstruction problems, for example, studying matrices instead of graphs:  Given some binary matrix M, suppose we are presented with the collection of its rows and columns in independent arbitrary orderings. From this information, are we able to recover the original matrix and will it be unique? We present an algorithm that identifies whether there is a unique ordering associated with a set of rows and columns, and outputs either the unique correct orderings for the rows and columns or the full collection of all valid orderings and valid matrices. We show that for matrices with entries that are i.i.d. Bernoulli(p), that for p >2log(n)/n that the matrix is indeed unique with high probability. This is a joint work with Caelan Atamanchuk and Luc Devroye.

Title: Sylvester-Gallai type configurations and Polynomial Identity Testing

Abstract: The classical Sylvester-Gallai theorem in combinatorial geometry says the following:

If a finite set of points in the Euclidean plane has the property that the line joining any two points contains a third point from the set, then all the points must be collinear. 

More generally, a Sylvester-Gallai (SG)-type configuration is a finite set of geometric objects with certain local dependencies. A remarkable phenomenon is that the local constraints give rise to global dimension bounds for linear SG-type configurations, and such results have found far reaching applications to complexity theory and coding theory. In particular, SG-type configurations have been extremely useful in applications to Polynomial Identity Testing (PIT), a central problem in algebraic complexity theory.

In this talk, we will discuss non-linear generalizations of SG-type configurations which consist of polynomials. We will discuss how uniform bounds on SG-configurations give rise to deterministic poly-time algorithms for the PIT problem. I’ll talk about results showing that these non-linear SG-type configurations are indeed low-dimensional as conjectured by Gupta in 2014. This is based on joint works with A. Garg, R. Oliveira and S. Peleg.

Title: The spectra of lift and factored lifts of graphs or digraphs

Abstract: In this talk, we first explain the path we took from defining polynomial matrices to a generalization of voltage graphs that we call combined voltage graphs. Moreover, we give a general definition of a matrix associated with a combined voltage graph, which allows us to provide a new method for computing the eigenvalues and eigenspaces of such graphs. 

Title: Sometimes you can't compute on encrypted data 

Abstract: A long dream in cryptography was "fully homomorphic encryption": the ability to perform computations on encrypted data. This way, a cloud provider can do your computations without compromising your privacy. The first general method was proposed by Gentry in 2009, but this method and follow up works are all too slow for many purposes. Naturally, we want to do better. This talk will explain some work in progress from a pessimistic perspective: when can we prove that faster methods are impossible.

Title: A Theory of Alternating Paths and Blossoms, from the Perspective of Minimum Length - Part 1

Abstract: It is well known that the proof of some prominent results in mathematics took a very long time --- decades and even centuries. The first proof of the Micali-Vazirani (MV) algorithm, for finding a maximum cardinality matching in general graphs, was recently completed --- over four decades after the publication of the algorithm (1980). MV is still the most efficient known algorithm for the problem. In contrast, spectacular progress in the field of combinatorial optimization has led to improved running times for most other fundamental problems in the last three decades, including bipartite matching and max-flow.

The new ideas contained in the MV algorithm and its proof remain largely unknown, and hence unexplored, for use elsewhere.

The purpose of this two-talk-sequence is to rectify that shortcoming.

Title: Introducing the interval poset associahedron

There will be a pre-seminar presenting relevant background at the beginning graduate level starting at 1pm.

Abstract: Given a permutation, we define its interval poset to be the set of all intervals ordered by inclusion. In this framework, a 'tube' is a convex connected subset, while a 'tubing' denotes a collection of tubes, that are pairwise either nested or disjoint. The interval poset associahedron is a polytope, whose faces correspond to proper tubes and whose vertices correspond to maximal tubings of the interval poset of a given permutation.

If we start with a simple permutation, the resulting interval poset associahedron will be isomorphic to the permutahedron. And if we consider inverse permutations, it turns out, that they yield identical associahedra.

If there is time, I will discuss another order on permutations, the Bruhat order, and compare it to the permutahedron.

Title: A Theory of Alternating Paths and Blossoms, from the Perspective of Minimum Length - Part 2

Abstract: It is well known that the proof of some prominent results in mathematics took a very long time --- decades and even centuries. The first proof of the Micali-Vazirani (MV) algorithm, for finding a maximum cardinality matching in general graphs, was recently completed --- over four decades after the publication of the algorithm (1980). MV is still the most efficient known algorithm for the problem. In contrast, spectacular progress in the field of combinatorial optimization has led to improved running times for most other fundamental problems in the last three decades, including bipartite matching and max-flow.

The new ideas contained in the MV algorithm and its proof remain largely unknown, and hence unexplored, for use elsewhere.

The purpose of this two-talk-sequence is to rectify that shortcoming.

Title: Infinite matroids on lattices

Abstract: There are at least two well-studied ways to extend matroids to more general objects - one can allow the ground set to be infinite, or instead define the concept of independence on a lattice other than a set lattice. I will discuss some nice ideas that arise when combining these two generalizations. This is joint work with Andrew Fulcher.