Events

Mathematics Colloquium

The first Mathematics Colloquium of the year will take place on Monday, September 21, at 4 pm in YR 321. The speaker will be Matheau Santana-Gijzen, a Ph.D. Candidate at the Johns Hopkins University. He will speak about Global guarantees for multimodal sampling:  Analytical tools and the value of advice.

Abstract. Sampling is a fundamental problem throughout statistics, the natural sciences and, increasingly,  artificial intelligence, with applications ranging from Bayesian posterior inference to generative modeling. Markov chain Monte Carlo (MCMC) methods provide a general framework for drawing samples from complex, high-dimensional distributions. However, these methods notoriously suffer when generating samples from multimodal distributions, as the underlying Markov process must cross regions of low probability mass to move between modes. Multimodal sampling thereby presents two distinct challenges: a sampler must explore each individual mode and it must also move probability mass between modes, maintaining the correct modal proportions. The first problem is local, while the second is global. 

In this talk, I will introduce analytical tools for decomposing the analysis of multimodal samplers into intra- and inter- modal parts and then show how these pieces can be combined to provide global guarantees. This perspective leads to a formalization of advice—the structural information provided to a sampler. I will present my work on two samplers: sequential Monte Carlo (SMC) and the reweighted annealed leap-point sampler (Re-ALPS), these samplers operate with two distinct forms of advice—no advice and weak advice. SMC demonstrates what guarantees can be achieved with minimal information, while Re-ALPS shows how prior information about mode locations can be used to preserve global proportions.  I will conclude by connecting these concepts back to rare-event simulation, controlled language generation and diffusion-based generative models. 

The next Mathematics Colloquium will take place on Thursday, October 15, at 4 pm in YR 321. The speaker will be Dr. Youran Sun, a Postdoctoral Associate at the University of Maryland at College Park. He will speak about Autonomous research: From prompt engineering to reinforcement learning.

Abstract. LLMs are making research production increasingly scalable. However, how to use LLMs to conduct research, how capable they are at research, and how to improve their research capabilities remain open questions. We first present Agon, a massively parallel, cross-disciplinary research pipeline, and discuss how it can reduce the time that human scientists spend on research. We then evaluate Agon on more narrowly scoped problems, including symbolic regression, alpha discovery, and PDE solving. Next, we discuss the limitations of LLM-based research, including which parts of the research process still require human involvement and how to construct benchmarks that expose the weaknesses of LLMs as research agents. Finally, I will discuss how to use RL to teach LLMs to conduct research. Together, these topics outline a path from prompting LLMs to perform research toward training them to become more capable research agents.

Recent Mathematics Colloquium Talks

Date	Topic	Speaker
Apr 29, 2026	Weyl, subconvex \( L^p \)-sets and arithmetic functions	Prof. Trevor D. Wooley,  Purdue University
Apr 22, 2026	Understanding the spline geometry of deep neural networks	Dr. Imtiaz Humayun,  Google Research
Mar 25, 2026	Weak assumptions in algebra and logic	Jacob Stern,  CUNY Graduate Center
Feb 25, 2026	Reducibility of convex cones	Dr. Michael Orlitzky, Viabit, LLC
Nov 10, 2025	Connecting noticing of student thinking to responsive teaching in math and science	Dr. Tara Barnhart, Chapman University
Oct 27, 2025	Introduction to statistical meta-analysis with an application	Dr. Bimal Sinha, UMBC
Oct 21, 2025	Large zeros of linear recurrence sequences	Dr. Florian Luca, Stellenbosch University
Apr 25, 2025	Defining connected components	Dr. Alfred Dolich, Kingsborough CC (CUNY)
Apr 24, 2025	Middle-school students’ meanings of points from quantitative and covariational reasoning perspectives	Dr. Halil Tasova, Cal State - San Bernardino
Mar 14, 2025	A multi-species, multi-stimuli rheological model for living polymers	Dr. Michael Cromer, RIT
Feb 13, 2025	Using data science to understand patterns of social harm	Dr. George Mohler, Boston College

Seminar Meetings

The Mathematics Seminar is the venue where Towson faculty and students report on their research activities. The next talk will occur on Monday, September 14, 2026, at 3:30 pm in YR 321. Dr. Diana Cheng will present a sabbatical talk on Increasing teachers’ self-efficacy with using artificial intelligence for culturally responsive mathematics lesson development.

Abstact. Our local district partners with our institution to enhance teacher learning by funding teachers to earn their Master of Science degrees in Mathematics Education.  In the first course of the masters degree sequence, teachers take a course entitled MTED 613: Mathematical Modeling in the School Classroom. Within the course, teachers revised existing mathematics lessons to become more culturally responsive. Frequently, they used artificial intelligence to assist in their background research on adapting the context of the lesson to our local region and for lesson planning ideas.  Pre- and post-surveys from Fall 2025 indicate that elementary school teachers increased in their self-efficacy of using artificial intelligence for instruction.  I describe several class sessions to explain the types of professional learning activities which helped the teachers gain self-efficacy.  I will also explain how I mentored MTED 613 students along with additional mathematics education students to present their culturally responsive mathematics lessons at local and national conferences in Spring and Summer 2026, including a conference funded by the National Science Foundation through the Northeast Noyce Regional Network which I co-hosted with TU’s Center for STEM Excellence and TU mathematics colleagues.

In addition, several research groups have research seminars in their respective research areas:

• ASRM Seminar (coordinated by Min Ji):
  Meets on Fridays at 10 am
• Number Theory Seminar (coordinated by A. Kumchev and N. McNew):
  Meets on Tuesdays at 4:00 pm in YR 321.

Recent and Upcoming Seminars

• April 28, 2026: Number Theory & Logic Seminar.
  Dr. Margaret Thomas, Purdue University, will give a talk.
• March 6, 2026: ASRM Speaker Series
  Matt Deveney '01', FSA, EA, presented a lecture How an actuary plays the "Game of Life."
• October 16, 2025: Number Theory Seminar.
  Dr. Spencer Hamblen, McDaniel College, gave at talk on Sums of k-th powers in ramified p-adic rings.
  
• September 16, 2025: Number Theory Seminar.
  Dr. Vefa Goksel gave a talk on Square patterns in dynamical orbits.

Recent Mathematics Seminar & Sabbatical Talks

• On Wednesday, May 13, 2026, Dr. Kristin Frank presented a sabbatical talk on Rethinking the Role of Quantitative Reasoning in Students' Mathematical Thinking.
• On Monday, April 13, 2026. Dr. Xiaoyin Wang presented a sabbatical lecture on Bayesian Analysis for Negative Binomial INGARCH Models with Minimal Prior Information. 
• On Friday, March 13, 2026, Dr. Mostafa Aminzadeh presented a sabbatical talk on Extending Approximate Bayesian Computation to Nonlinear Regression Models: The Case of Composite Distribution. 
• On Wednesday, March 4, 2026, Dr. Chris Cornwell presented a sabbatical lecture on the Functional Dimension of ReLU Neural Networks.
• On November 17, 2025, Dr. Vincent Guingona presented a sabbatical lecture on Applications of Model Theory to Statistical Learning Theory.
• On November 3, 2025, Dr. Kimberly Corum presented a sabbatical lecture on Leveraging Justice-Centered Making to Support Mathematics Teaching and Learning.
• On April 7, 2025, Dr. Banghee So, gave a seminar on A Novel Neural Network Model with Predictive Power and Interpretability for Insurance Pricing. 
• On March 24, 2025, Dr. Jing Tian presented a sabbatical lecture on Parameter Analysis in Continuous Data Assimilation for Various Turbulence Models.
• On February 24, 2025, Dr. Sebastian Calvo presented a seminar on the Waldschmidt Constant of Complex Reflection Groups.
• On October 16, 2024, Dr. Min Deng presented a sabbatical lecture on Bayesian Inference for the Loss Models via Mixture Priors. 
• On September 30, 2024, Dr. Melike Kara Atas presented a sabbatical lecture on Improving Pre-Service Teachers’ Mathematical Knowledge for Teaching Fraction Concepts.