Seminars and Colloquia by Series

Solving Inverse Problems on Networks: Graph Cuts, Optimization Landscape, Synchronization

Series
Applied and Computational Mathematics Seminar
Time
Monday, April 15, 2019 - 13:55 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Shuyang Ling – New York University –
Information retrieval from graphs plays an increasingly important role in data science and machine learning. This talk focuses on two such examples. The first one concerns the graph cuts problem: how to find the optimal k-way graph cuts given an adjacency matrix. We present a convex relaxation of ratio cut and normalized cut, which gives rise to a rigorous theoretical analysis of graph cuts. We derive deterministic bounds of finding the optimal graph cuts via a spectral proximity condition which naturally depends on the intra-cluster and inter-cluster connectivity. Moreover, our theory provides theoretic guarantees for spectral clustering and community detection under stochastic block model. The second example is about the landscape of a nonconvex cost function arising from group synchronization and matrix completion. This function also appears as the energy function of coupled oscillators on networks. We study how the landscape of this function is related to the underlying network topologies. We prove that the optimization landscape has no spurious local minima if the underlying network is a deterministic dense graph or an Erdos-Renyi random graph. The results find applications in signal processing and dynamical systems on networks.

Prime tropical ideals

Series
Algebra Seminar
Time
Monday, April 15, 2019 - 12:50 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Kalina Mincheva – Yale University –

Tropical geometry provides a new set of purely combinatorial tools, which has been used to approach classical problems. In tropical geometry most algebraic computations are done on the classical side - using the algebra of the original variety. The theory developed so far has explored the geometric aspect of tropical varieties as opposed to the underlying (semiring) algebra and there are still many commutative algebra tools and notions without a tropical analogue. In the recent years, there has been a lot of effort dedicated to developing the necessary tools for commutative algebra using different frameworks, among which prime congruences, tropical ideals, tropical schemes. These approaches allows for the exploration of the  properties of tropicalized spaces without tying them up to the original varieties and working with geometric structures inherently defined in characteristic one (that is, additively idempotent) semifields. In this talk we explore the relationship between tropical ideals and congruences to conclude that the variety of a prime (tropical) ideal is either empty or consists of a single point. This is joint work with D. Joó.

High-dimensional knots, and rho-invariants by Patrick Orson

Series
Geometry Topology Seminar Pre-talk
Time
Monday, April 15, 2019 - 12:45 for 1 hour (actually 50 minutes)
Location
Skiles 006
Speaker
Patrick Orson – Boston College –

I will give a brief survey of concordance in high-dimensional knot theory and how slice results have classically been obtained in this setting with the aid of surgery theory. Time permitting, I will then discuss an example of how some non-abelian slice obstructions come into the picture for 1-knots, as intuition for the seminar talk about L^2 invariants.

Stability and bifurcation analysis of the period-T motion of a vibroimpacting energy generator

Series
CDSNS Colloquium
Time
Monday, April 15, 2019 - 11:15 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
L. Serdukova – School of Mathematics, Georgia Institute of Technology

Stability and bifurcation conditions for a vibroimpact motion in an inclined energy harvester with T-periodic forcing are determined analytically and numerically. This investigation provides a better understanding of impact velocity and its influence on energy harvesting efficiency and can be used to optimally design the device. The numerical and analytical results of periodic motions are in excellent agreement. The stability conditions are developed in non-dimensional parameter space through two basic nonlinear maps based on switching manifolds that correspond to impacts with the top and bottom membranes of the energy harvesting device. The range for stable simple T-periodic behavior is reduced with increasing angle of incline β, since the influence of gravity increases the asymmetry of dynamics following impacts at the bottom and top. These asymmetric T-periodic solutions lose stability to period doubling solutions for β ≥ 0, which appear through increased asymmetry. The period doubling, symmetric and asymmetric periodic motion are illustrated by bifurcation diagrams, phase portraits and velocity time series.

Completely log-concave polynomials and matroids

Series
ACO Colloquium
Time
Friday, April 12, 2019 - 15:05 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Cynthia Vinzant – North Carolina State University, Raleigh, NC

Stability is a multivariate generalization for real-rootedness in univariate polynomials. Within the past ten years, the theory of stable polynomials has contributed to breakthroughs in combinatorics, convex optimization, and operator theory. I will introduce a generalization of stability, called complete log-concavity, that satisfies many of the same desirable properties. These polynomials were inspired by work of Adiprasito, Huh, and Katz on combinatorial Hodge theory, but can be defined and understood in elementary terms. The structure of these polynomials is closely tied with notions of discrete convexity, including matroids, submodular functions, and generalized permutohedra. I will discuss the beautiful real and combinatorial geometry underlying these polynomials and applications to matroid theory, including a proof of Mason’s conjecture on numbers of independent sets. This is based on joint work with Nima Anari, Kuikui Liu, and Shayan Oveis Gharan.

(*Refreshments available at 2:30pm before the colloquium.*)

Aubry-Mather theory for homeomorphisms

Series
Dynamical Systems Working Seminar
Time
Friday, April 12, 2019 - 15:05 for 1 hour (actually 50 minutes)
Location
Skiles 246
Speaker
Adrian P. Bustamante – Georgia Tech

In this talk we will follow the paper titled "Aubry-Mather theory for homeomorphisms", in which it is developed a variational approach to study the dynamics of a homeomorphism on a compact metric space. In particular, they are described orbits along which any Lipschitz Lyapunov function has to be constant via a non-negative Lipschitz semidistance. This is work of Albert Fathi and Pierre Pageault.

Milnor K-Theory

Series
Student Algebraic Geometry Seminar
Time
Friday, April 12, 2019 - 12:00 for 1 hour (actually 50 minutes)
Location
Skiles 006
Speaker
Stephen McKean – Georgia Tech –

Milnor K-theory is a field invariant that originated as an attempt to study algebraic K-theory. Instead, Milnor K-theory has proved to have many other applications, including Galois cohomology computations, Voevodsky's proof of the Bloch-Kato conjecture, and Kato's higher class field theory. In this talk, we will go over the basic definitions and theorems of Milnor K-theory. We will also discuss some of these applications.

Random Neural Networks with applications to Image Recovery

Series
Stochastics Seminar
Time
Thursday, April 11, 2019 - 15:05 for 1 hour (actually 50 minutes)
Location
Skiles 006
Speaker
Paul Hand – Northeastern University –
Neural networks have led to new and state of the art approaches for image recovery. They provide a contrast to standard image processing methods based on the ideas of sparsity and wavelets. In this talk, we will study two different random neural networks. One acts as a model for a learned neural network that is trained to sample from the distribution of natural images. Another acts as an unlearned model which can be used to process natural images without any training data. In both cases we will use high dimensional concentration estimates to establish theory for the performance of random neural networks in imaging problems.

Fractional coloring with local demands

Series
Graph Theory Seminar
Time
Thursday, April 11, 2019 - 15:00 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Tom Kelly – University of Waterloo

In a fractional coloring, vertices of a graph are assigned subsets of the [0, 1]-interval such that adjacent vertices receive disjoint subsets. The fractional chromatic number of a graph is at most k if it admits a fractional coloring in which the amount of "color" assigned to each vertex is at least 1/k. We investigate fractional colorings where vertices "demand" different amounts of color, determined by local parameters such as the degree of a vertex. Many well-known results concerning the fractional chromatic number and independence number have natural generalizations in this new paradigm. We discuss several such results as well as open problems. In particular, we will sketch a proof of a "local demands" version of Brooks' Theorem that considerably generalizes the Caro-Wei Theorem and implies new bounds on the independence number. Joint work with Luke Postle.

Optimal estimation of smooth functionals of high-dimensional parameters

Series
High Dimensional Seminar
Time
Wednesday, April 10, 2019 - 15:00 for 1 hour (actually 50 minutes)
Location
Skiles 006
Speaker
Vladimir Koltchinskii – Georgia Tech –

Please Note: We discuss a general approach to a problem of estimation of a smooth function $f(\theta)$ of a high-dimensional parameter $\theta$ of statistical models. In particular, in the case of $n$ i.i.d. Gaussian observations $X_1,\doot, X_n$ with mean $\mu$ and covariance matrix $\Sigma,$ the unknown parameter is $\theta = (\mu, \Sigma)$ and our approach yields an estimator of $f(\theta)$ for a function $f$ of smoothness $s>0$ with mean squared error of the order $(\frac{1}{n} \vee (\frac{d}{n})^s) \wedge 1$ (provided that the Euclidean norm of $\mu$ and operator norms of $\Sigma,\Sigma^{-1}$ are uniformly bounded), with the error rate being minimax optimal up to a log factor (joint result with Mayya Zhilova). The construction of optimal estimators crucially relies on a new bias reduction method in high-dimensional problems and the bounds on the mean squared error are based on controlling finite differences of smooth functions along certain Markov chains in high-dimensional parameter spaces as well as on concentration inequalities.

Pages