Seminars and Colloquia by Series

The challenge of accurate prediction of fluid motion

Series
School of Mathematics Colloquium
Time
Thursday, August 22, 2024 - 11:00 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
William LaytonUniversity of Pittsburgh

Over the last 40 years there have been great advances in computer hardware, solvers (methods for solving Ax=b and F(x)=0), meshing algorithms, time stepping methods, adaptivity and so on. Yet accurate prediction of fluid motion (for settings where this is needed) is still elusive. This talk will review three major hurdles that remain: ensemble simulations, time accuracy and model stagnation. Three recent ideas where numerical analysis can help push forward the boundary between what can be done and what can't be done will be described. This talk is based on joint work with many. It should be completely understandable by grad students with a basic PDE class.

Induction for 4-connected Matroids and Graphs (Xiangqian Joseph Zhou, Wright State University)

Series
Graph Theory Seminar
Time
Tuesday, July 23, 2024 - 15:30 for 1 hour (actually 50 minutes)
Location
Skiles 005
Speaker
Xiangqian Joseph ZhouWright State University

A matroid $M$ is a pair $(E, \mathcal{I})$ where $E$ is a finite set, called the {\em ground set} of $M$, and $\mathcal{I}$ is a non-empty collection of subsets of $E$, called {\em independent sets} of $M$, such that (1) a subset of an independent set is independent; and (2) if $I$ and $J$ are independent sets with $|I| < |J|$, then exists $x \in J \backslash I$ such that $I \cup \{x\}$ is independent. 

A graph $G$ gives rise to a matroid $M(G)$ where the ground set is $E(G)$ and a subset of $E(G)$ is independent if it spans a forest. Another example is a matroid that comes from a matrix over a field $F$: the ground set $E$ is the set of all columns and a subset of $E$ is independent if it is linearly independent over $F$. 

Tutte's Wheel and Whirl Theorem and Seymour's Splitter Theorem are two well-known inductive tools for proving results for 3-connected graphs and matroids. In this talk, we will give a survey on induction theorems for various versions of 4-connected matroids and graphs.   
 

Contact surgery numbers

Series
Geometry Topology Seminar
Time
Monday, July 22, 2024 - 14:00 for 1 hour (actually 50 minutes)
Location
Skiles 006
Speaker
Rima ChatterjeeUniversity of Cologne

A fundamental result in 3-dimensional contact topology due to Ding-Geiges tells us that any contact 3-manifold can be obtained via doing a surgery on a Legendrian link in the standard contact 3-sphere. So it's natural to ask how simple or complicated a surgery diagram could be for a given contact manifold? Contact surgery number is a measure of  this complexity. In this talk, I will discuss this notion of complexity along with some examples. This is joint work with Marc Kegel.

REU poster session

Series
Time
Thursday, July 18, 2024 - 11:00 for 3 hours
Location
Skiles Atrium
Speaker

The annual School of Math REU summer poster session will take place 11-2 on Thursday July 18th in the Skiles Atrium. We have a group of more than 20 students presenting projects on a variety of subjects (info for most of the projects available here). There will also be some light snacks and coffee etc. Come by and see the hard work that the students have done this summer; the students will certainly appreciate your interest!

Moduli of Fano varieties and K-stability

Series
Job Candidate Talk
Time
Tuesday, July 2, 2024 - 11:00 for 1 hour (actually 50 minutes)
Location
ONLINE
Speaker
Harold BlumUniversity of Utah

Algebraic geometry is the study of shapes defined by polynomial equations called algebraic varieties. One natural approach to study them is to construct a moduli space, which is a space parameterizing such shapes of a given type (e.g. algebraic curves). After surveying this topic, I will focus on the problem of constructing moduli spaces parametrizing Fano varieties, which are a class of positively curved complex manifolds that form one of the three main building blocks of varieties in algebraic geometry. While algebraic geometers once considered this problem intractable due to various pathologies that occur, it has recently been solved using K-stability, which is an algebraic definition introduced by differential geometers to characterize when a Fano variety admits a Kähler-Einstein metric.

Quantitative convergence analysis of dynamical processes in machine learning

Series
Dissertation Defense
Time
Tuesday, June 25, 2024 - 10:30 for 2 hours
Location
Skiles 006 and online
Speaker
Yuqing WangGeorgia Tech

Zoom link: https://gatech.zoom.us/j/6681416875?pwd=eEc2WEpxeUpCRUFiWXJUM2tPN1MvUT09

This talk focuses on analyzing the quantitative convergence of selected important machine learning processes, from a dynamical perspective, in order to understand and guide machine learning practices. More precisely, it consists of four parts: 1) I will illustrate the effect of large learning rates on optimization dynamics in a specific setup, which often correlates with improved generalization. 2) The theory from part 1 will be extended to a unified mechanism of several implicit biases in optimization, including edge of stability, balancing, and catapult. 3) I will concentrate on diffusion models, which is a concrete and important real-world application, and theoretically demonstrate how to choose its hyperparameters for good performance through the convergence analysis of the full generation process, including optimization and sampling. 4) The generalization performance of different architectures, namely deep residual networks (ResNets) and deep feedforward networks (FFNets), will be discussed.

On Extremal, Algorithmic, and Inferential Problems in Graph Theory

Series
Dissertation Defense
Time
Thursday, May 30, 2024 - 13:00 for 2 hours
Location
Skiles 005 and Online: https://gatech.zoom.us/j/6125656239
Speaker
Abhishek DhawanGeorgia Tech Math

In this dissertation we study a variety of graph-theoretic problems lying at the intersection of mathematics, computer science, and statistics. This work consists of three parts, all of which use probabilistic techniques. 

In Part 1, we consider structurally constrained graphs and hypergraphs. We examine a celebrated conjecture of Alon, Krivelevich, and Sudakov regarding vertex coloring. Our results provide improved bounds in all known cases for which the conjecture holds. We introduce a generalized notion of local sparsity and study the independence and chromatic numbers of graphs satisfying this property. We also consider multipartite hypergraphs, a natural extension of bipartite graphs. We show how certain probabilistic techniques for problems on bipartite graphs can be adapted to multipartite hypergraphs, and are therefore able to extend and generalize a number of results.

In Part 2, we investigate edge coloring from an algorithmic standpoint. We focus on multigraphs of bounded maximum degree, i.e., $\Delta(G) = O(1)$. Following the so-called augmenting subgraph approach, we design deterministic and randomized algorithms using a near-optimal number of colors in the sequential setting as well as in the LOCAL model of distributed computing. Additionally, we study list-edge-coloring for list assignments satisfying certain local constraints, and describe a polynomial-time algorithm to compute such a coloring.

Finally, in Part 3, we explore a number of statistical inference problems in random hypergraph models. Specifically, we consider the statistical-computational gap for finding large independent sets in sparse random hypergraphs, and the computational threshold for the detection of planted dense subhypergraphs (a generalization of the classical planted clique problem). We explore the power and limitations of low-degree polynomial algorithms, a powerful class of algorithms which includes the class of local algorithms as well as approximate message passing and power iteration.

Dynamic Stability in Stochastic Gradient Descent

Series
CDSNS Colloquium
Time
Friday, May 24, 2024 - 15:30 for
Location
Skiles 254
Speaker
Dennis ChemnitzFU Berlin

Please Note: Streaming via Zoom: https://gatech.zoom.us/j/91390791493?pwd=QnpaWHNEOHZTVXlZSXFkYTJ0b0Q0UT09

Most modern machine learning applications are based on overparameterized neural networks trained by variants of stochastic gradient descent. To explain the performance of these networks from a theoretical perspective (in particular the so-called "implicit bias"), it is necessary to understand the random dynamics of the optimization algorithms. Mathematically this amounts to the study of random dynamical systems with manifolds of equilibria. In this talk, I will give a brief introduction to machine learning theory and explain how almost-sure Lyapunov exponents and moment Lyapunov exponents can be used to characterize the set of possible limit points for stochastic gradient descent.

Validated enclosures of Fourier coefficients in Banach spaces of analytic functions

Series
CDSNS Colloquium
Time
Friday, May 10, 2024 - 15:30 for 1 hour (actually 50 minutes)
Location
Skiles 254
Speaker
Jean-Philippe LessardMcGill University

Please Note: Streaming available via Zoom: https://gatech.zoom.us/j/91390791493?pwd=QnpaWHNEOHZTVXlZSXFkYTJ0b0Q0UT09

This presentation introduces a methodology for generating computer-assisted proofs (CAPs) aimed at establishing the existence of solutions for nonlinear differential equations featuring non-polynomial analytic nonlinearities. Our approach combines the Fast Fourier Transform (FFT) algorithm with interval arithmetic and a Newton-Kantorovich argument to effectively construct CAPs. A key highlight is the rigorous management of Fourier coefficients of the nonlinear term Fourier series, achieved through insights from complex analysis and the Discrete Poisson Summation Formula. We demonstrate the effectiveness of our method through two illustrative examples: firstly, proving the existence of periodic orbits in the Mackey-Glass (delay) equation, and secondly, establishing the existence of periodic localized traveling waves in the two-dimensional suspension bridge equation.

This is joint work with Jan Bouwe van den Berg (VU Amsterdam, The Netherlands), Maxime Breden (École Polytechnique, France) and Jason D. Mireles James (Florida Atlantic University, USA)

Thermodynamic formalism for and hitting time statistics for random open dynamical systems

Series
CDSNS Colloquium
Time
Wednesday, May 8, 2024 - 14:00 for 1 hour (actually 50 minutes)
Location
Skiles 006
Speaker
Jason AtnipQueensland University

Please Note: In this talk we present some recent results on thermodynamic formalism for random open dynamical systems. In particular, we poke random holes in the phase space and prove the existence of unique equilibrium states on the set of surviving points as well as find the rate at which mass escapes through these holes. If we consider small holes, through a perturbative approach, we are able to make a connection to extreme value theory and hitting time statistics. Furthermore, we prove a Gumbel's law and show that the distribution of multiple returns to small holes is asymptotically compound Poisson distributed.

Pages