Multi-Fidelity Machine Learning for Uncertainty Quantification | Dr. S. De | JHU-IITD SMaRT Seminar

Описание: This talk is part of the Scientific Machine Learning Research Talks (SMaRT) Seminar Series, a joint initiative between Johns Hopkins University and IIT Delhi.

🔹 Speaker:
Dr. Subhayan De
Assistant Professor, Department of Mechanical Engineering
Northern Arizona University, USA

🔹 Talk Title:
Multi-Fidelity Machine Learning for Uncertainty Quantification

🔹 Date: Wednesday, September 10, 2025
Time: 7:30 PM IST | 2:00 PM GMT | 10:00 AM EDT

📄 Abstract

Uncertainty is inherent in engineering systems, arising from sources such as variability in material properties, incomplete knowledge of governing physics, and discretization errors. Capturing and propagating these uncertainties is essential for robust prediction, risk-aware decision-making, and design under uncertainty. Yet, traditional uncertainty quantification (UQ) methods like Monte Carlo simulations, can be prohibitively expensive for high-dimensional, nonlinear, and multiscale systems.

In this talk, Dr. De introduces a multi-fidelity machine learning framework that combines bi-fidelity data fusion, transfer learning, and neural operator models to address UQ in complex dynamical systems. The framework exploits low-cost, lower-fidelity models alongside sparse high-fidelity simulations to achieve computational efficiency without sacrificing accuracy. The key highlights of the talk are (a) Bi-fidelity DeepONets for partially known systems, (b) ℓ₁-regularized training for sparse and interpretable representations, (c) Transfer learning architectures robust to biased low-fidelity models. Together, these advances pave the way for scalable, reliable, and efficient UQ in scientific and engineering applications.

👤 About the Speaker

Dr. Subhayan De is an Assistant Professor in the Department of Mechanical Engineering at Northern Arizona University (NAU). He leads a research group dedicated to developing probabilistic, data-driven frameworks that combine machine learning with physics-based modeling for the design of multi-scale, multi-functional structural systems and materials under uncertainty.

He earned his Ph.D. in Civil Engineering (2018) and M.S. in Electrical Engineering (2016) from the University of Southern California (USC), supported by the Viterbi Ph.D. Fellowship and the Gammel Scholarship.

📌 Follow the SMaRT Seminar Series for more talks at the intersection of AI, physics, and uncertainty quantification.

#UncertaintyQuantification #MultiFidelity #ScientificMachineLearning #NeuralOperators #DeepONet #JHUIITD #SciML #RiskAwareAI