applied math

Joint Inverse Problems/Data Sciences/Applied Math Seminar at Colorado State University

Thursday 2:00-3:00PM, Weber 223

[Back to Department of Mathematics]
[Back to Organizer]











Zoom Meeting link  
Meeting ID: 883 4756 8032

Fall 2026

Sep 03   Sep 17   Oct 15   Oct 22   Oct 29 (Duo)   Nov 12 (Duo)   Dec 03  
 
 
 
Sep 03   Back to top

Yuanqing Wu

Independent Researcher

Title  From Computational Bottlenecks to Trustworthy AI in Scientific Computing

Abstract  Large-scale scientific simulations are often limited not by the absence of a governing model, but by computational kernels that are expensive and repeatedly evaluated across many grid cells, time steps, and parameter studies. This talk asks how such bottlenecks can be redesigned while preserving the conservation, stability, and physical consistency required for trustworthy scientific computation. Using subsurface flow as a demanding testbed, I will present three connected themes. First, I will show how locally adaptive sparse-grid surrogates move pressure-temperature flash calculations out of the online compositional-flow simulation. In the reported experiments, the surrogate accelerates the flash-calculation component by approximately 2,000 times, improves the total simulation time by up to 21 times, and reduces stored values by about 90 percent. Second, I will discuss a fully conservative mixed finite-element method for reactive dissolution and wormhole propagation. By treating fluxes as primary unknowns, the method enforces local mass balance and supports stability analysis and a priori error estimates, with numerical results exhibiting nearly first-order convergence. Third, I will describe a physics-informed neural surrogate for flash calculations that incorporates feasible-output constraints, material balance, phase equilibrium, and a physics-based phase decision. This removes the need to generate labeled flash data offline, while revealing an important accuracy challenge near phase-transition boundaries. I will conclude with a broader research program in which learning proposes computational actions, classical numerical methods verify them, and mathematical analysis provides guarantees for hybrid scientific AI.
 
 
 
Sep 17   Back to top

SIAM-IDA Panel on Summer Internship



 
 
 
Oct 15   Back to top

Emily King

Department of Mathematics, Colorado State University

Title  Symmetries in Neural Networks

Abstract  This talk will bring together two topics in machine learning---equivariant neural networks and superposition---under the umbrella of symmetry. No prior knowledge about neural networks will be assumed.

Although neural networks yield powerful results in many applications, there are drawbacks to their use. For example, there are many fields where the amount of labeled data needed to train neural networks is insufficient. Further, the internal decision-making process of a neural network is often opaque, leading to a desire for more interpretable techniques. These concerns have led some to consider leverage group actions to develop algorithms like equivariant neural networks. We will present theoretical results and numerical testing of a novel neural network structure that enforces partial invariance.

Neural networks have the ability to store more concepts than neurons; this is called superposition. Frames are redundant generalizations of bases which often have non-trivial symmetry groups. Frame theory has connections to combinatorial design theory, quantum information theory, and more. (In fact, a recent LLM-driven proof solving the biggest open problem in frame theory took down parts of Hilbert's 12 Problem and the Stark Conjectures in number theory.) Work done by practitioners investigating trained neural network models showed certain frame-like properties emerge. In this talk, we will some theoretical results inspired by trained models.  
 
 
Oct 22   Back to top

Stephen Dauphin

Sandia National Laboratories

Title 

Abstract   
 
 
Oct 29   Back to top

Dylan Soller

Department of Mathematics, Colorado State University

Title  Computing Body Surface Potentials on EIT Electrodes from Simulated Transmembrane Potential i Distributions in the Ventricular Myocardium

Abstract  The forward problem of electrocardiography (ECG) is to predict the voltage distribution on the surface of a body, based on a known electric source distribution in the heart. This forward problem is required to solve the corresponding inverse problem, which is to reconstruct a source distribution in the heart from voltage measurements on the body's surface. In this talk, we present a forward model pipeline that simulates ECG data collection on EIT electrodes, modelled after those on modern systems such as ACT5. The pipeline includes cardiac source generation using the monodomain based cardiac simulation software CardioMat, and subsequent propagation to body surface potentials via the finite element method.

Chris Rocheleau

Department of Mathematics, Colorado State University

Title  Anatomically-Informed 3-Dimensional Linearized Eit Reconstructions for Real-Time Monitoring of i Cardiopulmonary Function

Abstract  Real-time imaging is one of the strengths of EIT imaging. However, due to the ill-posedness of the inverse problem the images show less spatial resolution than other imaging methodologies. Further, many linearized methods perturb from a physiologically infeasible assumption that the patient is a cylinder of constant admittivity. In this talk we present the Mean 3-Dimensional Atlas Newton (M3AN) algorithm for thoracic EIT imaging. This algorithm extends the formulation of existing one-step Newton method solvers to use an anatomically-informed shape and admittivity distribution as an initial condition based on an Anatomical Atlas comprised of real-world CT scans. This results in images with improved resolution while still maintaining processing-time benefits of other one-step methods. Imaging and derived measures computed using this algorithm will be compared to existing algorithms for clinical data.  
 
 
Noc 12   Back to top

Shantam Gulati, Jennifer Meuller

Department of Mathematics, Colorado State University

Title 

Abstract   
 
 
Dec 03   Back to top

Stephen Becker

Department of Applied Mathematics, Univeristy of Colorado, Boulder

Title 

Abstract