AIcES Seminar | NguyenLab

Topology Meets XAI

Fri, 01 May 2026 14:30:00 -0400

Deep learning models are trained on massive datasets, yet their learned representations remain largely opaque. This talk explores how topological data analysis and visualization, particularly mapper graphs, can reveal structure within high-dimensional embedding spaces.

Dr. Bei Wang Phillips is an Associate Professor in the School of Computing, an Adjunct Associate Professor in Mathematics, and a faculty member of the Scientific Computing and Imaging Institute at the University of Utah. Her research spans topological data analysis, data visualization, and computational topology.

Network Reconstruction in Evolutionary Biology

Tue, 21 Apr 2026 16:00:00 -0400

A central challenge in biology and artificial intelligence is learning latent structure from data. In phylogenetics, this means reconstructing evolutionary histories from genetic data, often moving beyond trees to networks that capture hybridization and gene flow.

This talk focuses on identifiability: whether the underlying structure can be uniquely recovered from observed data. Computational algebraic geometry provides tools for understanding when network features can be recovered and when they cannot, even with unlimited data.

AI for Synthetic Biology

Fri, 27 Mar 2026 14:30:00 -0400

Synthetic biology aims to design novel or improved biological systems using engineering principles, with broad applications in medicine, chemicals, food, and agriculture. Because biological systems are complex, synthetic biology remains difficult to perform in a quantitative and predictive manner.

This talk highlights recent work on AI tools and AI-powered self-driving biofoundries that accelerate the design-build-test-learn cycle. Examples include ECNet for protein engineering, CLEAN for enzyme function prediction, EZSpecificity for enzyme substrate specificity prediction, generative AI for mitochondrial targeting sequences, and BioAutomata for protein, pathway, and metabolic engineering.

Dr. Huimin Zhao is the Steven L. Miller Chair of chemical and biomolecular engineering at UIUC, director of NSF AI Institute for Molecule Synthesis, NSF iBioFoundry, and NSF Global Center for Biofoundry Applications, and Editor in Chief of ACS Synthetic Biology.

Perception, Learning, and Memory in Brains and AI models: What Aligns and What Doesn't

Fri, 27 Feb 2026 14:30:00 -0500

This talk examines whether AI foundation models are converging on computational principles used by brains for perception, learning, and memory. It argues for moving beyond task accuracy toward mechanistic comparisons of representations and learning dynamics.

Dr. Zoran Tiganj is an Assistant Professor of Computer Science at Indiana University. His research combines artificial intelligence, cognitive science, and computational neuroscience with the objective of building artificial agents that learn from temporal and spatial regularities in the world.

AI discovery in Physics and Astronomy: Promise and Challenges

Fri, 06 Feb 2026 14:30:00 -0500

Can AI help answer known questions in physics and astronomy, and can it uncover phenomena scientists did not anticipate? With modern datasets growing faster than traditional modeling capacity, AI is becoming a partner in inference and discovery by extracting physical information, flagging anomalies, and suggesting new directions.

Dr. Cecilia Garraffo is founder and director of the AstroAI Institute at the Center for Astrophysics | Harvard & Smithsonian. Her work applies computational and deep learning techniques to stellar evolution, star-planet interactions, astrophysical simulations, and exoplanet atmosphere characterization.

Geometry-aware generative models for scientific data generation

Fri, 09 Jan 2026 14:30:00 -0500

Biological and biochemical data are increasingly high-dimensional and noisy, but they are often constrained to low-dimensional manifolds that encode meaningful state spaces, dynamics, and regulatory structure. This talk presents a unifying view of generative modeling for manifold-structured scientific data, focusing on methods that explicitly learn and exploit geometry rather than density alone.

The talk begins with SUGAR, a graph-based diffusion framework for geometry-aware data generation, then discusses PHATE, Neural FIM, GAGA, and RiTINI for manifold learning, Riemannian metrics, complex biological generative modes, and regulatory interaction network inference.

Topological Machine Learning: From Drug Discovery to Cancer Detection

Wed, 03 Dec 2025 14:30:00 -0500

Topological machine learning applies ideas from topology to extract stable, multiscale summaries of structure in complex data. This talk introduces core tools through intuition and simple visuals, then highlights applications in drug discovery and cancer detection from histopathology.

Baris Coskunuzer is a Professor of Mathematics at the University of Texas at Dallas and leads the Topological Machine Learning Group. His research bridges geometry, topology, and artificial intelligence, with recent work on graph learning and medical image analysis.

Statistical Foundations of Deep Generative Models

Fri, 07 Nov 2025 14:30:00 -0500

Generative AI has achieved remarkable performance in many domains, motivating theoretical work on its foundations. This talk examines deep generative models from a nonparametric distribution-estimation perspective, where the generator is parameterized by a deep neural network.

Lizhen Lin is a professor of statistics in the Department of Mathematics at the University of Maryland and director of the statistics program. Her expertise includes Bayesian modeling, high-dimensional theory, statistics on manifolds, network analysis, and foundations of deep neural network models.

How AI Learned to Talk and What It Means

Wed, 01 Oct 2025 12:30:00 -0400

Christopher Summerfield is Professor of Cognitive Neuroscience at the University of Oxford and a Research Director at the UK AI Security Institute. His research connects cognitive science, neuroscience, and artificial intelligence, focusing on principles of human learning and decision-making.

An Overview of the Use of Large Language Models with Medical Data

Fri, 05 Sep 2025 14:30:00 -0400

This talk discusses the use of large language models for extracting social determinants of health data from clinical text, identifying patients at high risk of radiation therapy interruption, and analyzing high-dimensional metabolic data.

Dr. Robert Davis is the University of Tennessee-Oak Ridge National Laboratory Governor’s Chair for Biomedical Informatics at UTHSC, Professor of Pediatrics, and founding director of the Center for Biomedical Informatics.