Theme

Models meet Data

Computational and data-enabled science has become the third pillar of science, completing theory and experimentation. Approaches within computational and data-enabled science may be categorized as model-based and data-driven. On the one hand, mathematical and physical models are powerful tools in modern data science to discover, analyze, and predict underlying mechanisms of real-world applications. The classical modeling approach was and still is driving many discoveries. Data-driven scientific approaches with the help of artificial intelligence and tools from machine learning, on the other hand, have advanced and influenced our daily lives in recent years. Current state-of-the-art scientific methods merge both approaches to enhance discoveries through physics-informed and data-driven hybrid approaches.

Important Dates

Date Event
March 1 Application deadline
June 16 On-campus REU phase begins
August 8 Last day of on-campus phase

Projects

Models Meet Data

Joint Analysis of X-ray Ptychography and X-ray Fluorescence Reconstruction
Development of a joint optimization framework that simultaneously reconstructs X-ray ptychography and fluorescence data, leveraging their complementarity for improved image reconstruction compared to separate methods.
Dec 16, 2025
Hybrid Regularization for Random Feature Models
This project investigates hybrid LSLU, a novel iterative regularization method for training random feature models that avoids the double descent phenomenon while requiring fewer iterations and less computational cost than existing approaches like hybrid LSQR.
Aug 1, 2025
Inverse problems and uncertainty quantification for the control of dynamical systems
Developed a complete pipeline for optimal control of predator-prey systems under uncertainty. Combined Bayesian parameter estimation with MCMC, pseudo-spectral methods, and Pontryagin’s Maximum Principle to propagate uncertainty through control decisions for Lotka-Volterra dynamics.
Aug 1, 2025
Model Aware and Data Driven Inference
Exploring linear encoder-decoder networks for interpretable inverse problem solutions using Bayes risk minimization.
Jul 1, 2025

Cohort Activities

To facilitate interactions across the teams and interactions between participants and faculty, our site emphasizes cohort activities. Those activities include:

  • a weekly project meeting with breakfast
  • a weekly seminar with lunch
  • ad-hoc seminars on background material (e.g., optimization, machine learning, PDEs, …)
  • social events
  • mid-term presentations
  • final poster session with poster awards
  • many mentoring and advising opportunities (e.g., grad school and career panels, …)

We involve participants in planning these activities and adapt them based on their background knowledge and project needs.