Colloquium Series 2021-22

This Colloquium Series is sponsored by the Mathematics and Computer Science Department.

Unless otherwise noted, all talks begin at 4:30 pm in Jepson Hall 109. Masks are required to attend a talk.

For more information, contact the colloquium chair, Dr. Michael Kerckhove.

Upcoming:

March 28: Jin Lu, Ph.D., Assistant Professor, Computer and Information Science, University of Michigan-Dearborn

Topic: Predictive Modelling via Substructure-based Machine Learning

Abstract: In the recent decade, machine learning has been substantially developed and has demonstrated great success in various domains such as web search, computer vision, and natural language processing. Despite of its practical success, many of the applications involve solving complex problems based on building one end-to-end model, which often neglect to analyze whether structural information is well investigated. In this talk, it shows that if a certain sub-structure occurs in sample data, it is possible to solve the related problem with lower computational cost and higher accuracy than that of the classic machine learning methods. We propose to employ two granular data structures, e.g., task bi-sectioning, or multi-task clustering to design new statistical models for two learning problems respectively.

April 18: Andre Shannon, Computer Science Honors Presentation, R’22

Topic: Developing Provably Robust Explanation Methods for Image Classifiers

Abstract: Machine learning systems can use up to billions of parameters, making it difficult to understand a model’s “reasoning” behind its decisions. An explanation method seeks to better understand this “reasoning”, usually for specific inputs. However, recent work shows that these explanation methods are not robust, slight changes in the input can lead to completely different explanations. In addition, attacks have been developed that exploit this and fool both the machine learning classifier and the explanation method. This talk explores our efforts to develop provably robust explanation methods for image classifiers

Past Talks:

March 14: Van Nall, Ph.D., Professor of Mathematics, University of Richmond

Topic: Dynamics of Shift Maps on Inverse Limits with a Single Set Valued Function on [0,1]

Abstract: A set valued function on [0,1] is just a relation on [0,1]. That is it is just a subset [0,1] x [0,1]. We will look only at closed subsets of [0,1]x[0,1]. For historical reasons we prefer to think of this closed set as a set value function from the second coordinates of the points in [0,1] x [0,1] into the closed subset of [0,1]. For example, for each in [0,1]. The inverse limit of such a function is the set of all sequences of the form where for each The standard way to think about such sequences is that they are points in the Hilbert Cube. This collection of points/sequences in the Hilbert cube can be a fascinating and rich structure and the function you get by throwing away the first term of the sequence and shifting the remaining terms to the left by one has incredible dynamical properties. We are talking about chaos, positive entropy, shift maps on Cantor sets, and all of that kind of thing. A lot of chaos from a very simple closed set in [0,1] x [0,1] is the what this game is all about. We will also indicate how this approach might be used to answer an important open question in dynamics.

February 21: Peter K. Johnson, R’12, Ph.D. Candidate at University of Virginia

Topic: A Brief Introduction to Knot Theory

Abstract: The study of knots in 3-space has a rich mathematical history and is a very active area of current mathematical research. To study and classify different types of knots, one often assigns to each knot some quantity (for example a number, a polynomial, a vector space, etc.) which remains unchanged if you wiggle and deform the knot. Such a quantity is called a knot invariant. In this talk, I will describe two fundamentally important knot invariants: the Jones polynomial discovered by Vaughan Jones in 1984 and the Alexander polynomial discovered by James Waddell Alexander II in 1923. No prior background in knot theory will be assumed other than some geometric intuition. I will also allocate some time to discuss my experience as an undergraduate student at the University of Richmond and my subsequent experience as a math graduate student.

February 7: Martha Whamond, W'17, Health Industries Advisory Senior Associate

Topic: Careers in Analytics: Transforming Healthcare with Data-Driven Decisions

Abstract: If you’ve ever wondered about a career in analytics - whether you plan on diving right into the job market or pursuing a graduate degree - this seminar is for you. The booming data analytics industry is expected to quadruple in the next 5 years and the demand for quantitative skills is greater than ever. Despite the growing opportunities, students often wonder what a career in data analytics really entails and where to start the job search. To help answer these questions, Martha will discuss her career in consulting and how she leverages her background in mathematics to help healthcare organizations make data-driven decisions.

November 1: Oscar Javier Chaparro Arenas, Assistant Professor of Computer Science, College of William & Mary

Title: Getting a Ph.D. (and an M.Sc.): Myths and Facts

Abstract: A Ph.D. in Computer Science (CS) can open doors to multiple career opportunities in academia and industry, such as corporate research and development (R&D) jobs, faculty positions, high-tech startups, and senior-level product development, to name a few. However, enrolling in a graduate program is an impactful decision with life-changing consequences, and as a result, must be taken carefully, with as much information as possible. CS juniors and seniors may have numerous questions regarding a Ph.D. (and an M.Sc.): “Why should I get a Ph.D.?”, “How will a Ph.D. help my career?”, “Where should I get my Ph.D. from?”, "How are my Ph.D. studies going to be funded?", “Do I need an M.Sc. degree for getting a Ph.D.?”, “What’s the difference between an M.Sc. and a Ph.D.?”, etc.

October 25: Dr. Eric Swartz, Associate Professor of Mathematics, College of William & Mary

Topic: Covering Numbers of Rings with Unity

Abstract: Given an algebraic structure (group, ring, etc.), a cover is defined to be a collection of proper substructures (e.g., subgroups, subrings, etc.) whose set theoretic union is the whole structure. Assuming such an algebraic structure has a cover, its covering number is defined to be the size of a minimum cover. I will discuss the rich history of this problem as well as recent joint work with Nicholas Werner on the covering number of a ring with unity. No prior knowledge will be assumed beyond the basic definitions of groups and rings.

October 4: Dr. Paul Kvam, Professor of Statistics, University of Richmond Note-Moved to Jepson Hall 118

Topic: Careers in Statistics

Abstract: If you have considered a career in Statistics and Data Science, you should attend this seminar by Professor Paul Kvam to find out what opportunities are out there for University of Richmond graduates. We will learn about classes and opportunities offered at UR that can help you in your pursuits. The best positions require a graduate degree, so we will discuss how to pick out the right graduate program and improve your chances for gaining admittance to a top school.

September 27: Dr. Marcella Torres, Director of Mathematical Studies, University of Richmond

Title: A Machine Learning Method for Parameter Estimation and Sensitivity Analysis

Abstract: Dr. Torres will discuss the application of a supervised machine learning method, decision tree algorithms, to perform parameter space exploration and sensitivity analysis on mathematical models. Decision trees can provide complex decision boundaries and can help visualize decision rules in an easily digested format. This aids in understanding the predictive structure of a dynamic model and the relationship between input parameters and model output. She will illustrate how this process can be used in the early stages of model development for a simple ordinary differential equation model of HIV dynamics.

August 30: Student Summer Research Presentations

If you are interested in viewing recordings of the research presentations, contact Dr. Kerckhove.

August 31: Student Summer Research Presentations

If you are interested in viewing recordings of the research presentations, contact Dr. Kerckhove.