Events

Mon. September 30, 2024 - Youssef Marzouk - Massachusetts Institute of Technology(MIT)

Professor of Aeronautics and Astronautics

Co-director, Center for Computational Science and Engineering

Massachusetts Institute of Technology

Transport methods for Bayesian inference and optimal experimental design

3:45 pm (refreshments 3:45 pm, seminar begins at 4:00 pm)

RTH 526

Abstract: Measure transport has emerged as a versatile tool in probabilistic modeling and inference, offering a unifying perspective on various computational challenges. This talk explores the core principles of transport maps and their ability to induce couplings between probability measures, facilitating efficient simulation and analysis. We will survey the diverse landscape of transport representations, from polynomials and invertible neural networks to ODE flow maps, highlighting how these constructions capture different notions of low-dimensional structure in probabilistic models.

The presentation will then focus on recent advancements in two key areas:

1. Nonlinear ensemble filtering: This talk explores novel transport-based algorithms that generalize the ensemble Kalman filter to nonlinear settings, offering improved performance in challenging filtering problems.

2. Simulation-based inference: We will investigate how transport maps can be leveraged to enhance the efficiency and accuracy of inference in scenarios where only forward simulations are available.

Additionally, this talk explores the application of transport-based density estimates in bounding information-theoretic objectives for optimal experimental design, demonstrating the broad utility of this framework in decision-making under uncertainty.

Bio: Youssef Marzouk is the Breene M. Kerr (1951) Professor of Aeronautics and Astronautics at the Massachusetts Institute of Technology (MIT), and co-director of the Center for Computational Science and Engineering within the MIT Schwarzman College of Computing. He is also a core member of MIT's Statistics and Data Science Center and a PI in the MIT Laboratory for Information and Decision Systems (LIDS).

His research interests lie at the intersection of statistical inference, computational mathematics, and physical modeling. He develops new methodologies for uncertainty quantification, Bayesian computation, and machine learning in complex physical systems, motivated by a broad range of engineering and science applications. His recent work has centered on algorithms for inference, with applications to data assimilation and inverse problems; dimension reduction methodologies for high-dimensional learning and surrogate modeling; optimal experimental design; and transportation of measure as a tool for inference and stochastic modeling.

He received his SB, SM, and PhD degrees from MIT and spent four years at Sandia National Laboratories before joining the MIT faculty in 2009. He is also an avid coffee drinker and an occasional classical pianist.

Thur. October 17, 2024 - Jessica Zhang - Carnegie Mellon University

George Tallman Ladd and Florence Barrett Ladd Professor

Department of Mechanical Engineering

Courtesy Appointment in Biomedical Engineering

Carnegie Mellon University

From neurological disorders to additive manufacturing: integrating isogeometric analysis
with deep learning and digital twins

9:00 am (refreshments 9 am, seminar begins at 9:30 am)

RTH 526

Abstract: For neurological disorders, a novel phase field model is introduced, simulating neurite outgrowth and disorders using IGA. Combining IGA with convolutional neural networks, the talk analyzes key parameters affecting neurodevelopmental disorders and presents a PDE-constrained optimization model for neurodegenerative disorders. Additionally, an IGA-based physics-informed graph neural network is developed to predict intracellular transport in complex neuron geometries.

This talk explores the integration of physics-based simulations with data-driven modeling, focusing on isogeometric analysis (IGA), deep learning, and digital twins for two main applications: neurological disorders and additive manufacturing (AM). In the AM domain, the talk covers AI-driven inverse design for 4D printing, IGA-based topology optimization for heat exchangers, and rapid geometry distortion prediction in metal printing processes. Ongoing efforts include developing digital twins for efficient process control in laser powder bed fusion (LPBF) manufacturing.

Bio: Jessica Zhang is the George Tallman Ladd and Florence Barrett Ladd Professor of Mechanical Engineering at Carnegie Mellon University, with a courtesy appointment in Biomedical Engineering. She earned her B.Eng. in Automotive Engineering and M.Eng. in Engineering Mechanics from Tsinghua University, and her M.Eng. in Aerospace Engineering and Ph.D. in Computational Engineering and Sciences from The University of Texas at Austin. Her research interests include computational geometry, isogeometric analysis, the finite element method, data-driven simulations, and image processing, with a strong focus on their applications in computational biomedicine and engineering. Zhang has co-authored over 230 publications in peer-reviewed journals and conference proceedings and is the author of the book Geometric Modeling and Mesh Generation from Scanned Images (CRC Press). Her work spans both theoretical development and practical applications, contributing significantly to advancements in both fields. She is a Fellow of prominent societies, including ASME, SIAM, IACM, USACM, IAMBE, AIMBE, SMA, and ELATES at Drexel, highlighting her distinguished reputation in the field. Currently, she serves as the Editor-in-Chief of Engineering with Computers, further establishing her leadership in computational science and engineering research.

Tue. October 22, 2024 - Vikram Gavini - University of Michigan, Ann Arbor

Department of Mechanical Engineering, Department of Materials Science and Engineering

Professor, Mechanical Engineering
Professor, Materials Science and Engineering

Large-scale electronic structure calculations of extended defects in materials

9:00 am (refreshments 9 am, seminar begins at 9:30 am)

RTH 526

Abstract:

Defects play a crucial role in influencing the macroscopic properties of solids—examples include the role of dislocations in plastic deformation, dopants in semiconductor properties, and domain walls in ferroelectric properties. These defects are present in very small concentrations (few parts per million), yet, produce a significant macroscopic effect on the materials behavior through the long-ranged elastic and electrostatic fields they generate. Notably, the strength and nature of these fields, as well as other critical aspects of the defect-core are all determined by the electronic structure of the material at the quantum-mechanical length-scale. However, carefully converged electronic structure studies on extended defects, such as dislocations, have been out of reach due to the cell-size and periodicity limitations of the widely used electronic structure codes.

This talk will discuss the recent developments that have enabled large-scale density functional theory (DFT) calculations, paving the way for electronic structure studies of defects. The first part of the talk will discuss the development of computational methods and numerical algorithms for conducting fast and accurate large-scale DFT calculations using adaptive finite-element discretization, which form the basis for the recently released DFT-FE open-source code. The second part of the talk will focus on electronic structure studies of dislocations using the developed methods and the insights obtained into fundamental questions such as: What is the core size of a dislocation? Are forces on dislocations solely from elastic interactions? Recent studies on using DFT-FE to understand the energetics of <c+a> dislocations in Mg, and the energetics and nucleation kinetics of quasicrystals (ScZn7.33) will be discussed.