ACM Seminar
The Applied and Computational Mathematics (ACM) seminar series since 2020.
Click HERE for the schedule.
The power of least squares in the worst-case and learning setting
Speaker: Felix Bartel (Technische Universität Chemnitz)
Least squares approximation is a time-tested method for function approximation based on samples. It is a natural to compare it to approximations based on arbitrary linear functionals as a benchmark. In this talk we will present recent results from the information-based complexity community showing the optimality of the least squares algorithm. We will consider the worst-case setting: drawing points which are good for a class of functions and the learning setting: we approximate an individual function based on possibly noisy samples. We support our findings with numerical experiments.
Time: February 24, 2023 2:30pm-3:30pm
Location: LeConte 440
Leveraging computational modeling to understand biomedical diseases
Speaker: Wenrui Hao (Pennsylvania State University)
In this talk, I will explore the use of computational and mathematical modeling as critical tools in understanding and predicting the progression of biomedical diseases. I will present two recently developed modeling approaches, pathophysiology-driven modeling and data-driven modeling, and provide examples of each. For pathophysiology-driven modeling, I will introduce a mathematical model of atherosclerosis and discuss how it provides a personalized cardiovascular risk by solving a free boundary problem. This model also presents some interesting mathematical challenges that can deepen our understanding of cardiovascular risk. For data-driven modeling, I will use Alzheimer's disease as an example to illustrate the idea of learning a mathematical model from clinical data when the pathophysiology of a disease is not well understood. I will also discuss how this approach has been applied to personalized treatment studies of the recently FDA-approved Alzheimer's medication, aducanumab.
Time: March 3, 2023 2:30pm-3:30pm
Location: LeConte 440
Host: Yi Sun
Modeling plug formation in films inside tubes: impact of surfactant, viscosity stratification, and slip
Speaker: Reed Ogrosky (Virginia Commonwealth University)
Viscous liquid films coating the interior of a tube occur in a variety of applications. If the film is thick enough, it may pinch off and form a plug, occluding the tube. In this talk I will discuss recent work examining the impact of surfactant, slip, and viscosity stratification on plug formation in a model for film flow. Implications for understanding occlusion in human airways will be discussed.
Time: February 17, 2023 2:30pm-3:30pm
Location: LeConte 440
Host: Paula Vasquez
Machine learning of self organization from observation
Speaker: Ming Zhong (Illinois Institute of Technology)
Self organization (aka collective behaviors) occurs naturally in crystal formation, cell aggregation, social behaviors, etc. It is challenging and intriguing to understand self organization from the mathematical point of view. We offer a statistical/machine learning approach to explain self organization from observation data; moreover, our learning approach can aid in validating and improving the modeling of self organization. We develop a learning framework to derive physically meaningful dynamical systems to understand self organization from observation. We then investigate the steady state properties of our learned estimators. We also extend the learning approach for dynamical models constrained on Riemannian manifolds. We further improve our learning capability to infer interaction variables as well as interaction kernels. We study the effectiveness of our learning method on the NASA Jet Propulsion Laboratory's modern Ephemerides. Upon careful inspection of our model, we discover that it even captures potion of the general relativity effects. A complete learning theory on second-order systems is presented, as well as two new models on emergence of social hierarchy and concurrent emergence of flocking and synchronization.
Time: February 10, 2023 2:30pm-3:30pm
Location: LeConte 440
Host: Siming He and Changhui Tan
On the problem of emergence arising in hydrodynamic systems of collective behavior
Speaker: Roman Shvydkoy (University of Illinois Chicago)
Emergence is a phenomenon of formation of collective outcomes in systems where communications between agents has local range. In dynamics of swarms such outcomes often represent a globally aligned flock or congregation of aligned clusters. The classical result of Cucker and Smale states that alignment is unconditional in flocks that have global communication with non-integrable radial tails. Proving a similar statement for purely local interactions presents a major mathematical challenge. In this talk we will overview three programs of research directed on understanding the emergent phenomena: hydrodynamic topological interactions, kinetic approach based on hypocoercivity, and spectral energy method. We present a novel framework based on the concept of environmental averaging which allows us to obtain coercivity estimates leading to new flocking results.
Time: January 27, 2023 2:30pm-3:30pm
Location: LeConte 440
Host: Changhui Tan
Quantitative steepness, semi-FKPP reactions, and pushmi-pullyu fronts
Speaker: Jing An (Duke University)
We will discuss the algebraic structure of a large class of reaction-diffusion equations and use it to study the long-time behavior of the solutions and their convergence to traveling waves in the pulled and pushed regimes, as well as at the pushmi-pullyu boundary. A new quantity named as the shape defect function is introduced to measure the difference between the profiles of the solution and the traveling waves. In particular, the positivity of the shape defect function, combined with a new weighted Hopf-Cole transform and a relative entropy approach, plays a key role in the stability arguments. The shape defect function also gives a new connection between reaction-diffusion equations and reaction conservation laws at the pulled-pushed transition. This is joint work with Chris Henderson and Lenya Ryzhik.
Time: November 18, 2022 2:30pm-3:30pm
Location: LeConte 205
Host: Siming He
On the construction of 3D incompressible Euler equilibria by magnetic relaxation
Speaker: Federico Pasqualotto (Duke University)
Magnetic relaxation is a conjectured general procedure to obtain steady solutions to the incompressible Euler equations by means of a long-time limit of an MHD system. In some regimes, the magnetic field is conjectured to “relax” to a steady state of the 3D Euler equations as time goes to infinity.
In this talk, I will first review the classical problem of magnetic relaxation, connecting it to questions arising in topological hydrodynamics. I will then present a general construction of steady states of the incompressible 3D Euler equations by a long-time limit of a regularized MHD system. We consider the so-called Voigt regularization, and our procedure yields non-trivial equilibria on the flat 3D torus and on general bounded domains.
This is joint work with Peter Constantin.
Time: October 28, 2022 2:30pm-3:30pm
Location: LeConte 205
Host: Siming He
A new approach to the mean-field limit of Vlasov-Fokker-Planck equations
Speaker: Pierre-Emmanuel Jabin (Pennsylvania State University)
We introduce a novel approach to the mean-field limit of stochastic systems of interacting particles, leading to the first ever derivation of the mean-field limit to the Vlasov-Poisson-Fokker-Planck system for plasmas in dimension 2 together with a partial result in dimension 3. The method is broadly compatible with second order systems that lead to kinetic equations and it relies on novel estimates on the BBGKY hierarchy. By taking advantage of the diffusion in velocity, those estimates bound weighted L p norms of the marginals or observables of the system uniformly in the number of particles. This allows to treat very singular interaction kernels between the particles, including repulsive Poisson interactions. This is a joint work with D. Bresch and J. Soler.
Time: December 2, 2022 2:30pm-3:30pm
Location: Virtually via Zoom
Host: Changhui Tan
A measure perspective on uncertainty quantification
Speaker: Amir Sagiv (Columbia University)
In many scientific areas, deterministic models (e.g., differential equations) use numerical parameters. Often, such parameters might be uncertain or noisy. A more honest model should therefore provide a statistical description of the quantity of interest. Underlying this numerical analysis problem is a fundamental question - if two "similar" functions push-forward the same measure, would the new resulting measures be close, and if so, in what sense? We will first show how the probability density function (PDF) of the quantity of interest can be approximated. We will then discuss how, through the lense of the Wasserstein-distance, our problem yields a simpler and more robust theoretical framework.
Finally, we will take a steep turn to a seemingly unrelated topic: the computational sampling problem. In particular, we will discuss the emerging class of sampling-by-transport algorithms, which to-date lacks rigorous theoretical guarantees. As it turns out, the mathematical machinery developed in the first half of the talk provides a clear avenue to understand this latter class of algorithms.
Time: November 11, 2022 2:30pm-3:30pm
Location: Virtually via Zoom
Host: Wolfgang Dahmen
Hadamard-Babich ansatz for point-source Maxwell's equations
Speaker: Jianliang Qian (Michigan State University)
We propose a novel Hadamard-Babich ansatz consisting of an infinite series of dyadic coefficients (three-by-three matrices) and spherical Hankel functions for solving point-source Maxwell's equations in an inhomogeneous medium so as to produce the so-called dyadic Green's function. Using properties of spherical Hankel functions, we derive governing equations for the unknown asymptotics of the ansatz including the travel time function and dyadic coefficients. By proposing matching conditions at the point source, we rigorously derive asymptotic behaviors of these geometrical-optics ingredients near the source so that their initial data at the source point are well-defined. To verify the feasibility of the proposed ansatz, we truncate the ansatz to keep only the first two terms, and we further develop partial-differential-equation based Eulerian approaches to compute the resulting asymptotic solutions. Numerical examples demonstrate that our new ansatz yields a uniform asymptotic solution in the region of space containing a point source but no other caustics.
Time: October 21, 2022 2:30pm-3:30pm
Location: Virtually via Zoom
Host: Lili Ju
More...
An overview of augmented strategy and applications
Speaker: Zhilin Li (North Carolina State University)
Considering the different backgrounds of the audience, I would like to present an overview of an augmented strategy for solving PDEs hoping to find more applications of the approach. The purpose of the augmented strategy is to decouple some complex systems, rescale or preconditioning PDEs. The augmented strategy makes it possible to obtain accurate and stable discretization. The idea of the augmented strategy for a complicated problem is to introduce some augmented variable(s) along a codimension on a manifold, like a boundary integral method of the source and/or dipole strengths except that no Green's function is needed, more flexible in terms of PDEs (linear or nonlinear), boundary conditions and source terms.
Some important applications will be discussed including the treatment of pressure boundary conditions (not free variables) in Stokes and Navier-Stokes equations; rescaling and fast algorithms for interface problems with large jump ratios, a fluid flow and Darcy's coupling in which the governing equations are different in different regions; and ADI methods for parabolic interface problems, and scattering problems modeled by Maxwell equations, and solver PDEs on irregular domains.
Time: September 30, 2022 2:30pm-3:30pm October 7, 2022 2:30pm-3:30pm
Location: Virtually via Zoom
Host: Qi Wang
Spectral renormalizations methods in physics
Speaker: Ziad Musslimani (Florida State University)
In this talk we shall outline a new method to solve initial and boundary value problems of physical relevance. The idea is to use the underlying physics (such as conservation laws or dissipation rate equations) combined with a dynamic renormalization process to numerically compute ground and excited states as well as time-dependent solutions. We will apply the method on a prototypical problems that arise in physics such as Gross-Pitaevski equation and Hartree-Fock.
Time: April 22, 2022 2:30pm-3:30pm
Location: Virtually via Zoom
Host: Qi Wang
Global solutions of quasi-geostrophic shallow water front problems
Speaker: Qingtian Zhang (West Virginia University)
In this talk, I will introduce the vortex front problem for quasi-geostrophic shallow water equation, which is also known as Hasegawa-Mima equation in plasma science. The contour dynamic equation of the vortex front will be derived, which is a nonlocal, nonlinear dispersive equation. The existence of global solutions will be proved when the initial data is small.
Time: March 25, 2022 2:30pm-3:30pm
Location: Virtually via Zoom
Host: Changhui Tan
How Math and AI are revolutionizing biosciences
Speaker: Guowei Wei (Michigan State University)
Mathematics underpins fundamental theories in physics such as quantum mechanics, general relativity, and quantum field theory. Nonetheless, its success in modern biology, namely cellular biology, molecular biology, biochemistry, genomics, and genetics, has been quite limited. Artificial intelligence (AI) has fundamentally changed the landscape of science, technology, industry, and social media in the past few years and holds a great future for discovering the rules of life. However, AI-based biological discovery encounters challenges arising from the structural complexity of macromolecules, the high dimensionality of biological variability, the multiscale entanglement of molecules, cells, tissues, organs, and organisms, the nonlinearity of genotype, phenotype, and environment coupling, and the excessiveness of genomic, transcriptomic, proteomic, and metabolomic data. We tackle these challenges mathematically. Our work focuses on reducing the complexity, dimensionality, entanglement, and nonlinearity of biological data in AI. We have introduced evolutionary de Rham-Hodge, persistent cohomology, persistent Laplacian, and persistent sheaf theories to model complex, heterogeneous, multiscale biological systems and thus significantly enhance AI's ability to handle biological datasets. Using our mathematical AI approaches, my team has been the top winner in D3R Grand Challenges, a worldwide annual competition series in computer-aided drug design and discovery for years. Using over two million genomes isolates from patients, we discovered the mechanisms of SARS-CoV-2 evolution and transmission and accurately forecast emerging SARS-CoV-2 variants.
Time: April 15, 2022 2:30pm-3:30pm
Location: Virtually via Zoom
Host: Qi Wang