UNCW Interdisciplinary Research Seminar Series 2025-2026: Healthcare in the Era of Machine Learning
For the past two decades, public and private organizations have heavily invested in machine learning due to increased availability of large datasets, fueling research and discovery across disciplines, especially healthcare. This trend has been amplified with the public popularity of large language models (LLM) and other generative artificial intelligence (AI) applications. In healthcare, machine learning has the potential to enhance predictive healthcare models, improve preventative care with real time monitoring, transform patient outcomes, facilitate drug discovery, and make healthcare more accessible and cost-effective.
This seminar series will bring a diverse range of scholars and experts, from nationally renowned doctors to applied mathematicians, from policy influencers to business founders, to embark on an enriched research journey with UNCW’s faculty and students.
Title: Uncertainty Quantification in Neural Networks with Applications to MRI Processing
Radu Balan, PhD
University of Maryland, College Park
- October 3rd, 2025 Friday: , @Congdon 1018 (Auditorium)
Title: TBD
Hau-Tieng Wu, M.D., Ph.D
New York University
Sep 12th Balan Bio
Dr. Balan is a Professor of Applied Mathematics at the University of Maryland. He is currently the program director of Applied Mathematics, Statistics and Scientific Computing program. Dr. Balan has extensive research in the theory of application of neural network and deep learning. He is Co-editor in chief of Applied and Computational Harmonic Analysis, a top journal in applied mathematics. Before joining Maryland, Dr. Balan was a senior research scientist at Siemens. He obtained a PhD in Computational and Applied Mathematics in Princeton University in 1998.
abstract:
In this talk we study Lipschitz properties of neural networks. In practical numerical examples (such as Alex Net, and scattering networks), estimations of local Lipschitz bounds are compared to these theoretical bounds. Based on the Lipschitz bounds, we next establish concentration inequalities for the output distribution with respect to a stationary random input signal. Such a Lipschitz analysis is next applied to medical image processing. Image reconstructions involving neural networks (NNs) are generally non-iterative and computationally efficient. However, without analytical expression describing the reconstruction process, the computation of noise propagation becomes difficult. Automated differentiation allows rapid computation of derivatives without an analytical expression. In this talk, the feasibility of computing noise propagation with automated differentiation was investigated. The noise propagation of image reconstruction by End-to-end variational-neural-network was estimated using automated differentiation and compared with Monte-Carlo simulation. The root-mean-square error (RMSE) map showed great agreement between automated differentiation and Monte-Carlo simulation over a wide range of SNRs.
