Virginia Tech® home

Colloquium

  • Colloquium
  • STAT 5924, CRN 92847
  • Fridays
  •  2:30 pm - 3:30 pm
  • 300 Seitz Hall

Colloquium Schedule Fall 2026

Join us for snacks, drinks, and conversation as we get to know each other better. We’ve reserved the bar area, so we’ll be comfortably inside if it’s too warm/wet outside. We’ll go around and introduce ourselves. Graduate students and faculty, please be ready to share: Your 5-year goal, your current research interests, and a hobby you enjoy

The Maroon Door

418 N Main St, Blacksburg, VA 24060

Wilson Wright

Bio: Dr. Wilson Wright is an Assistant Professor in the Department of Statistics and Data Science at the University of Missouri. Prior to joining MU, he received a PhD in Statistics from Colorado State University. His research focuses on developing spatial models for analyzing ecological and environmental data. In particular, his work considers new ways to model the distributions of plant and animal species while accounting for spatial dependencies and various forms of observation errors. His paper on “Continuous-space occupancy models” received the 2025 Best Paper in Biometrics by an International Biometric Society member Award.

Title: Continuous-space occupancy models

Abstract: Occupancy models are commonly used to infer species distributions over large spatial extents while accounting for imperfect detection. Current approaches, however, are unable to model species occurrence over continuous spatial domains while accounting for the discrete spatial domain of the observed data. We develop a new class of spatial occupancy models that embeds a change of spatial support between the observed data and occurrence process. We use a clipped Gaussian process to represent species occurrence in continuous space which can provide inferences at a finer resolution than the observed occupancy data. Our approach is beneficial because it allows for more realistic models of species occurrence, can account for species occurring in only a portion of a surveyed site, and can relate detection probabilities to these within-site occurrence proportions. We show how our model can be fit using Bayesian methods and develop a computationally efficient MCMC algorithm. We also show how our framework can be extended to handle multiple years of data to learn about species occupancy dynamics in continuous space. We demonstrate our model using simulated data and compare our approach to alternative spatial occupancy models. We also use our model to analyze ovenbird occurrence data collected in New Hampshire, USA.

Location: 300 Seitz 

Dr. Hongtu Zhu

Bio: Dr. Hongtu Zhu is the Kenan Distinguished Professor of Biostatistics, Statistics, Radiology, Computer Science and Genetics at the University of North Carolina at Chapel Hill. He was a DiDi Fellow and Chief Scientist of Statistics at DiDi Chuxing between 2018 and 2020 and held the Endowed Bao-Shan Jing Professorship in Diagnostic Imaging at MD Anderson Cancer Center between 2016 and 2018. He is an internationally recognized expert in statistical learning, medical image analysis, precision medicine, biostatistics, artificial intelligence, and big data analytics. He received an established investigator award from the Cancer Prevention Research Institute of Texas in 2016, the INFORMS Daniel H. Wagner Prize for Excellence in Operations Research Practice in 2019, the IMS 2027 Medallion award and Lecture, and the COPSS 2025 Snedecor Award. He has published more than 365 papers in top journals, including Nature, Science, Cell, Nature Genetics, Nature Communication, PNAS, AOS, JASA, Biometrika, and JRSSB, as well as presenting 65+ conference papers at top conferences, including meetings for Neurips, ICLR, ICML, AAAI, STAI-X,  and KDD. He is the coordinating editor of JASA and the editor of JASA ACS.

Title: Understanding CNN Efficiency: Statistical Generative Models for Unstructured Image Data

Abstract: Convolutional Neural Networks (CNNs) are foundational in modern image analysis due to their ability to efficiently learn feature representations. However, theoretical understanding of their efficiency remains limited, largely due to inadequate modeling of image structures and their interaction with CNNs. To address this, we introduce novel statistical generative models (SGMs) that decompose images into task-relevant signals and noise, capturing the complexities of natural image data. Based on these SGMs, we propose a feature mapping approach (FMA) to characterize the transformation from raw image data to feature vectors. We analyze CNNs' approximation capabilities, their adaptation to low-dimensional structures, and their efficiency in vision tasks, ultimately developing statistical learning theories for CNN-based image analysis. Our findings reveal the challenges inherent in vision tasks and highlight CNNs' remarkable efficiency in addressing them, providing new insights into their theoretical and practical capabilities. This is based on the joint work with Dr. Guohao Shen.

Location: 300 Seitz 

Simon Mak

Bio: Simon Mak is an Associate Professor in the Department of Statistical Science at Duke University. He works broadly on developing novel Bayesian ML/AI methods, theory, and algorithms for advancing science and engineering. His research is motivated by and applied to ongoing interdisciplinary collaborations in nuclear and high-energy physics, aerospace engineering, bioengineering, and public policy.

Title: ML/AI-accelerated autonomous experimentation with digital twins: a statistician's playground

Abstract: An emerging use of ML/AI in science is autonomous scientific experimentation, which has shown promise for self-driving laboratories in, e.g., material and biotech engineering. This automation is fueled by the rise of "digital twin" simulators that enable reliable virtual experimentation of complex physical systems. One bottleneck, however, is the time-intensive nature of virtual experiments, which can take thousands of CPU hours to perform on a supercomputer. In this talk, we explore several ways that ML/AI can accelerate autonomous experimentation for timely scientific decision-making. The first is the use of physics-informed ML/AI models that incorporate known physics, e.g., boundary information, to enhance model learning from experimental data. The second is the use of ML/AI-enhanced optimization solvers to effectively design experimental runs for complex high-dimensional systems. Time permitting, the third is the use of robust transfer learning models that effectively transfer information from existing literature (e.g., extracted via large language models) to guide experimentation. We explore the effectiveness of these approaches in ongoing scientific collaborations.

Location: 145 Steger

Taps Maiti

Bio:  Taps Maiti is a Research Foundation Distinguished Professor in the Department of Statistics and Probability at Michigan State University. He recently served as Program Director in the Division of Mathematical Sciences at the National Science Foundation, where he oversaw the Statistics Program, the Mathematical Foundations of Artificial Intelligence program, and interdisciplinary initiatives at the interface of mathematical sciences and biomedical research. He also serves as Special Advisor for AI Strategy and Partnerships in MSU's College of Natural Science.

His research spans statistical learning, Bayesian and computational statistics, high-dimensional and complex data, uncertainty quantification, and the statistical foundations of artificial intelligence. He is a Fellow of the American Statistical Association, the Institute of Mathematical Statistics, and the American Association for the Advancement of Science, and has held editorial leadership roles for several major statistics journals.

Professor Maiti has mentored 40 Ph.D. students and has held several academic leadership positions at Michigan State, including Graduate Director and Co-Director of the Center for Business and Social Analytics. His current research increasingly focuses on the statistical foundations of deep learning and artificial intelligence, with an emphasis on uncertainty, Bayesian methodology, scalable computation, and applications to complex data.

Title:  Statistical Foundations of Deep Learning: Toward Efficient and Trustworthy AI

Abstract:  Deep learning has transformed modern artificial intelligence through increasingly flexible models, large datasets, and scalable computation. For statisticians, however, its rapid development raises fundamental questions about statistical modeling, inference, theory, and computational efficiency.

In this talk, we present a statistical perspective on deep learning, focusing on Bayesian neural networks, variational inference, theoretical validation, and structurally sparse deep networks. We discuss how statistical modeling and optimization can provide theoretical guarantees, quantify approximation effects, and produce more efficient deep-learning models without substantial loss of predictive performance. Applications to high-dimensional and complex data illustrate the role of statistical methodology in developing reliable and computationally efficient AI.

More broadly, the talk explores how statistical theory, machine learning, and optimization can work together to build more principled and trustworthy AI.

Location: 300 Seitz 

Christian Johnson
Lisa Lupinacci
Brooke Marshall

A Panel Discussion: Statistics Careers and Research Opportunities in the Pharmaceutical Industry

Panelists:

  • Christian Johnson — AstraZeneca
  • Lisa Lupinacci — Merck
  • J. Brooke Marshall — Lilly

Panelists will be joining remotely

Christian Johnson, Senior Director Safety Health Environmental, AstraZeneca

Christian Johnson is an executive risk management professional, focused specifically in safety, health, environmental (SHE) and sustainability. He currently serves as Global Senior Director, Safety, Health, Environment & Sustainability at AstraZeneca, a leading global pharmaceutical company. His responsibilities span multiple regions including Europe, US, and Latin America. Christian's prior roles include 19 years of global and executive SHE leadership at General Electric, along with roles in the high technology, packaging, and automotive sectors. His experience ranges across diverse operations including manufacturing, construction, and field services throughout all business cycles—from startup and rapid expansion to acquisition integration, restructuring, and environmental remediation.

Christian is particularly interested in technology applications, including the use of Generative AI solutions to improve risk identification and reduction. He holds B.S. and M.S. degrees in Industrial Engineering from Virginia Tech. Christian is a Certified Industrial Hygienist, and a Certified Safety Professional. In spare time he loves boating on the Chesapeake Bay and endurance running, having completed multiple marathons including Boston. He, his wife, and cat Reeses reside in Howard County, Maryland. Their twin sons attend New York University.

Lisa Lupinacci, Ph.D., Senior Vice President, Biostatistics and Research Decision Sciences, Merck

Lisa Lupinacci is the Senior Vice President of Biostatistics and Research Decision Sciences (BARDS) at Merck. In her 29 years at Merck, Lisa has followed a unique career path that includes positions in 3 therapeutic areas and 3 functional areas (statistics, clinical research, project leadership). Since January 2022, Lisa has served as the head of BARDS, an organization that provides statistical, epidemiologic, health economic modeling, and results interpretation and publication expertise to support Merck’s product development programs from discovery through reimbursement. BARDS professionals work in the US, China, Brussels, London, and Zurich. Lisa is a member of the cross-industry Biopharmaceutical Statistical Leaders Consortium and a member of the American Statistical Association.

Lisa holds a Bachelor’s degree in mathematics from Villanova University and M.S. and Ph.D. degrees in Statistics from Virginia Tech.

Brooke Marshall, Ph.D., Senior Director of Development Statistics, Eli Lilly and Company

Brooke Marshall, Ph.D., is a Senior Director of Development Statistics at Eli Lilly and Company, where Brooke leads the synthetic molecule development statistics team. With nearly two decades of industry experience, Brooke provides statistical leadership and CMC support across process development, analytical methods, and regulatory submissions.

Previously, Brooke established the CMC Statistics department at Moderna as a Fellow, supporting process development, analytical validation, specifications and shelf-life setting, and regulatory submissions, on vaccines for COVID, RSV, and cancer. In earlier roles at GSK and Merck she supported scientists in bioassay development, stability, process validation, and late-stage clinical biostatistics, including statistical leadership on GARDASIL regulatory submissions.

Brooke holds B.S., M.S., and Ph.D. degrees in Statistics from Virginia Tech. As a dedicated mentor and educator, Brooke co-founded Moderna’s R&D Statistics Center of Excellence and has taught statistics to scientists throughout her career in industry.

Title: Careers and Research Opportunities in the Pharmaceutical Industry: A Panel Discussion 

  • Christian Johnson (AstraZeneca)
  • Lisa Lupinacci (Merck)
  • J. Brooke Marshall (Lilly)

Abstract: The pharmaceutical industry offers statisticians opportunities to contribute across the entire lifecycle of medical products, from early discovery and preclinical research through clinical trials, regulatory decision-making, manufacturing, and post-market surveillance. At the same time, advances in data science, artificial intelligence, and precision medicine continue to create new methodological challenges and research opportunities.

This panel brings together VT alumni from three pharmaceutical companies to discuss their career journeys, current roles, and perspectives on the evolving landscape of the industry. Panelists will share insights into organizational structures, emerging workforce needs, and the skills most valued in today's pharmaceutical companies. They will also discuss pressing statistical and data science challenges facing the industry and opportunities for collaboration between academia and industry. The session will conclude with an open question and answer period.

Location: 300 Seitz 

Lizhen Lin

Bio: Lizhen Lin is a Professor of Statistics in the Department of Mathematics at the University of Maryland.  Some of her recent interests are in understanding theoretical foundations of deep neural network models and AI from the lens of statistical theory.  Her other research expertise lie in (1) developing Bayesian model and theory for dimensional and infinite-dimensional models, and variational Bayes theory, (2) statistics on manifolds or broadly geometry & statistics, (3) statistical network analysis, and (4) geometric deep learning. Lizhen's applied interests lie in leveraging topological and geometric approaches for data analysis, particularly through the use of topological data analysis tools to study data spanning continuous and discrete structures, including DNA sequences and networks.

Title: Early-stopped aggregation: Adaptive inference with computational efficiency

Abstract: When considering a model selection or, more generally, an aggregation approach for adaptive statistical inference, it is often necessary to compute estimators over a wide range of model complexities including unnecessarily large models even when the true data-generating process is relatively simple, due to the lack of prior knowledge. This requirement can lead to substantial computational inefficiency. In this work, we propose a novel framework for efficient model aggregation called the early-stopped aggregation (ESA): instead of computing and aggregating estimators for all candidate models, we compute only a small number of simpler ones using an early-stopping criterion and aggregate only these for final inference. Our framework is versatile and applies to both Bayesian model selection, in particular, within the variational Bayes framework, and frequentist estimation, including a general penalized estimation setting. We investigate adaptive optimal property of the ESA approach across three learning paradigms. We first show that ESA achieves optimal adaptive contraction rates in the variational Bayes setting under mild conditions. We extend this result to variational empirical Bayes, where prior hyperparameters are chosen in a data-dependent manner. In addition, we apply the ESA approach to frequentist aggregation including both penalization-based and sample-splitting implementations, and establish corresponding theory. As we demonstrate, there is a clear unification between early-stopped Bayes and frequentist penalized aggregation, with a common "energy" functional comprising a data-fitting term and a complexity-control term that drives both procedures. We further present several applications and numerical studies that highlight the efficiency and strong performance of the proposed approach.

Location: 300 Seitz 

Location: 300 Seitz 

Download Schedules from Colloquia Gone by

Contact Information

Department of Statistics (MC0439)
Hutcheson Hall, RM 406-A, Virginia Tech
250 Drillfield Drive
Blacksburg, VA 24061

Phone: 540-231-5657

Department Head:
Robert B. Gramacy