Richmond Home

2018–19 Colloquium Series

This Colloquium Series is sponsored by the Math and Computer Science Department. Unless otherwise noted, all talks take place in Jepson Hall Room 109 at 4:30 p.m. Starting at 4 p.m. light refreshments will be served in the lounge outside of Jepson 212 — please join us. Talks will be added as they are scheduled; please check back often.

Up Next

April 15: Raymond Cheng, Old Dominion University, Department of Mathematics & Statistics

Title: A Fun Exercise in Probability

Abstract: We'll look at several dramatically different approaches to solving a simple problem involving coin tosses.

April 22: Honors Students take Center Stage. Join us as computer science honors students Tuan Le, David Qin, and Hanglin "Jojo" Zhou present their theses.

Past Events

April 1: Greg Morrisett, Cornell, Dean of Computing and Information Sciences

Title: Securing Software through Proof Engineering

Abstract: The computers upon which we all depend, from laptops to servers to cell phones to embedded controllers, run software that is full of bugs.  Today, attackers find it relatively easy to exploit these bugs to gain access and control of these computers.  Is there anything we can do to substantially change this landscape?  One promising approach is based on machine-checked proofs of safety, correctness, and/or security.  I’ll discuss some of the recent research advances that show why we think this is a promising approach, and what challenges still remain.

March 20 Special We’ve invited a speaker under the auspices of the Integrated and Inclusive Science Program (co-sponsored by Dept. of Geography and the Environment) Location: Gottwald Science Center

Samson Hagos, Earth Scientist, Pacific Northwest National Lab

Dr. Hagos’ research interests are focused on understanding and modeling of precipitation processes over a wide range of spatio-temporal scales, from the life-cycles of individual convective cells, to tropical intra-seasonal oscillations, atmospheric rivers and inter-annual to multi-decadal variations of monsoon systems.

Title: South Asian monsoon rainfall in climate models: a link between inter-model spread and the representations of tropical convection

Abstract: The Coupled Model Intercomparison Project provides a community-based infrastructure that allows scientists to analyze and improve their models. Phase five of the project (CMIP5) involves/involved over 20 different climate models from research centers around the world. Collectively, the mean value of the simulated Asian monsoon rainfall amounts from current models is less than the actual measured precipitation (dry bias) and variability across models (inter-model spread) is larger than anticipated or desired. The origins of the bias and spread have not been well understood. Using moisture and energy budget analysis that exploits the weak temperature gradients in the tropics, we show that the bias and the inter-model spread are related to the non-linear relationship between precipitation and precipitable water (total moisture in the atmospheric column) found in previous observational studies. The effect of the non-linearity is to amplify the spread among models with high precipitable water. This behavior is also reflected in the projected response of the models to global warming. Our study shows the response of the South Asian Monsoon to warming, as well as the uncertainty in the projected response to global warming, are likely over-estimated because of the outsized contribution of models with high precipitable water over the tropics.

March 18: Larry Leemis, 2019 Gaines Chair in Mathematics and Professor of Mathematics at the College of William & Mary

Title: A Probability Calculator, Football Field Position, and Confidence Intervals

Abstract:  This talk considers three topics in stochastic modeling. The first is using a computer algebra system to perform probability calculations.  The second is how to visualize a mixed type random variable (part discrete and part continuous), illustrated by the field position in football after a kickoff.  The third considers ongoing work in finding a confidence interval for the Bernoulli parameter and finding a confidence region for the parameters in a two-parameter univariate probability distribution.

March 4: Emily Dodwell, AT&T Labs

Title: From Theory to Practice: A Machine Learning Use Case for Advertising at AT&T

Abstract: Emily Dodwell, a Senior Inventive Scientist in the Data Science and AI Research Organization at AT&T Labs, will present a recent project to develop a machine learning-based media targeting strategy for television advertising campaigns. Emily will discuss the computational challenges inherent in the scale of training data, potential solutions her team considered to tackle the business problem, as well as theoretical intuition for the final two machine learning algorithms they chose to compare for implementation.

January 28: Laura Ellwein-FixVirginia Commonwealth University, Department of Mathematics

Title: Parameter Identifiability of a Respiratory Mechanics Model in an Idealized Preterm Infant

Abstract:  The complexity of mathematical models describing respiratory mechanics has grown in recent years to integrate with cardiovascular models and incorporate nonlinear dynamics, but has rarely been studied in the context of patient-specific observable data. This study investigates parameter identification of a previously developed nonlinear respiratory mechanics model tuned to the physiology of 1 kg preterm infant, using local deterministic sensitivity analysis, subset selection, and gradient-based optimization. The model consists of 4 differential state equations with 34 parameters to predict airflow and dynamic pulmonary volumes and pressures generated under six simulation conditions. The relative sensitivity solutions were calculated with finite differences and a sensitivity ranking was created for each parameter and simulation. Subset selection found independent parameters that could be estimated for all six simulations. The combined analysis produced a subset of 6 independent sensitive parameters identifiable with observable data. Optimizations performed using pseudo-data with perturbed nominal parameters estimated parameters within 5% of nominal values on average, demonstrating the feasibility of studying patient-specific infant data with these methods.

November 26: Neal Bushaw, Virginia Commonwealth University, Department of Mathematics

Title: Turán Numbers and their Variants

Abstract:  Among the oldest questions in extremal graph theory lies a gem: For an n-vertex graph, how many edges can a graph possibly have while avoiding a particular subgraph? This problem dates back to the 1930s, and when it was answered (for complete graphs) by Pál Turán in the 1940s, the 'Turán Number' was born; given a graph G and a natural number n, we define the Turán Number as the maximum number of edges among all n vertex graphs with no subgraph isomorphic to G.  This problem not only has application within graph theory, but within other areas of mathematics and science.

In this general audience talk we'll talk about the history of graph theory in general, and this question specifically, as well a simple variation which leads to my own research ('What if instead of forbidding all copies of G, we allow one or two or ten?'). No graph theory background will be assumed -- I'll start by defining a graph, and build everything from there.

November 14: David Mimno, Cornell University, Information Science, Computing and Information Science **Special Wednesday Session**

Title: Putting the Data in Data Science

Abstract: "One of the most powerful conceptual tools in computing is abstraction. If you can recognize a class of problems that all share the same form, you can apply the same solution over and over. But this same power is also dangerous. We are tempted to put all of our attention on algorithms and treat data sets as interchangeable. I will describe several case studies in which small variations in input data can have surprisingly large impacts on the resulting outputs. I argue that data care -- far from being a trivial or menial task -- is often the most impactful part of a data science process."

November 12: Juraj Foldes, University of Virginia, Department of Mathematics

Title: Statistical Solutions of Differential Equations

Abstract: Many mathematical models possess very complicated or chaotic dynamics with solutions being extremely sensitive to parameters. In such situation, it is not feasible to follow one solution, but it is more practical to look at statistical properties of solutions. Famous complex systems arise in fluid dynamics, where two dimensional turbulent flows for large Reynold’s numbers can be approximated by solutions of incompressible Euler’s equation. As time increases, the solutions of Euler’s equation are increasing their disorder; however, at the same time, they are limited by the existence of infinitely many invariants. Analogously as the equilibrium statistical states are obtained in thermodynamics, we assume that the dynamics tend to limit profiles which maximize an entropy given the values of conserved quantities. These profiles, described by methods of Statistical Mechanics, are solutions of non-usual variational problems with infinite number of constraints. We will show how to analyze the problem and we will derive symmetry properties of entropy maximizers on symmetric domains. This is a joint work with Vladimír Šverák (University of Minnesota).

November 9: Evgenia Smirni, College of William and Mary, Computer Science Department ** Special Friday Session in Business School Room 114 at 4:00 PM **

Title: Getting a PhD in Computer Science: Myths and Facts

Abstract: A PhD in Computer Science can open doors to incredible career opportunities in academia or industry, such as corporate university R&D jobs, faculty positions, hi-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 with as much information as possible. In fact, CS juniors and seniors may have numerous questions regarding a PhD: “Why should I get a PhD?”, or “How will a PhD help my career?”, or “Where should I get my PhD from?”, or "How are my Ph.D. studies going to be funded?". This talk is designed to especially answer such questions.

This talk will provide students with critical information on getting a graduate degree in CS, and the benefits of doing so at William & Mary. First, I will describe what a CS PhD entails, and demystify the myths and facts about getting a PhD. I will also discus the factors one must consider when selecting a PhD program. I will then provide information on William & Mary (W&M), and specifically, the research strengths of the Computer Science department at W&M. Towards the end, students will have an idea on what it is like to get a PhD in CS, and specifically, what it would be like to get a PhD at W&M CS.

November 5: Doron Levy, University of Maryland, Department of Mathematics

Title:The Role of the Immune Response in Chronic Myelogenous Leukemia

Abstract: Targeted drugs have significantly improved treatment of chronic myelogenous leukemia (CML). Yet, most patients are not cured for undetermined reasons. In this talk we will describe our work on modeling the immune response to CML, the goal being to harness the immune response to better improve therapies. Along the way, we will discuss some our results on cancer vaccines, drug resistance, and cancer stem cells.  We will also emphasize the necessity of integrating  a mixed bag of mathematical tools in order to address complex biological problems.

October 29: Lincoln Mullen, George Mason University, Department of History & Art History **LOCATION CHANGED TO JEPSON 118**

Title: Finding Biblical Quotations in Historical Newspaper Corpora

Abstract: "America's Public Bible" is interactive work of digital scholarship that identifies quotations of the Bible in U.S. newspapers. This talk will explain how the project works from a computational perspective, including the challenges of finding quotations, working with historical corpora, and creating the website. It will also discuss how those computational methods connect to research questions in American religious history and religious studies. The site enables a disciplined serendipity that turns up new and unusual examples that one would be unlikely to find with more traditional methods, but which also rigorously contextualizes those examples. This method helps researchers make better evaluations of whether the phenomena under study are typical or unusual.

October 22: Zerotti Woods, University of Georgia, Department of Mathematics

Title: The Effects of Ill Conditioning in Neural Network

Abstract: "Deep Neural Networks have shown much success in solving problems in a diverse set of applications (i.e. computer vision, computational biology, finance, etc). Although we have proof about universal approximation of these networks the problem of training them is known to be very difficult. The ill conditioning of the hessian has been shown to be one of the sources of this difficulty. In this talk we will discuss problems and neural network architecture that causes a ill conditioned hessian. I will also discuss how this can interplay with analysis of high frequency telemetry data taken from a malaria infection on Non Human Primates."

October 8: Dominique Guillot, University of Delaware, Department of Mathematical Sciences

Title: Positivity Preserver Problems

Abstract:  "Determining which transformations map the set of positive semidefinite matrices into itself is a classical problem that continues to attract a lot of attention. I will give a historical account of matrix positivity and of operations that preserve it, and will discuss several applications of positivity preservers in geometry, combinatorics, and statistics. The talk should be accessible to anyone with a basic knowledge of linear algebra and calculus."

September 10: Student Summer Research Presentations II

Stephen Owen, Berke Nuri, Abhishek Shilpakar (Doug Szajda, mentor); Sophie Borchart, Palmer Robins, Jonathan Rodriguez (Jory Denny, mentor); Caleb Brooks, Aaqil Zakarya (Jory Denny, mentor); Basil Arafat (Jory Denny, mentor); Jojo Zhou (Jory Denny, mentor); Michael Bonifonte (Lewis Barnett, mentor); Hammed Hassan (Lewis Barnett, mentor)

September 3: Student Summer Research Presentations I

Diksha Kataria, Xinxuan Zhang, Shiyi Wang, Alamby He (Paul Kvam, mentor); Salar Ather (Taylor Arnold, mentor); Shuzhi Zeng & Nayzaw Aung Win (Lester Caudill, mentor); Maxine Xin (Prateek Bhakta, mentor); Sinan Kivanc (Prateek Bhakta, mentor); Miles Clikeman (Heather Russell, mentor)