Seminars

View all Seminars  |  Download ICal for this event

Hybrid Machine Learning with Data and Scientific Knowledge

Series: Department Seminar

Speaker: Dr. Priyabrata Saha, Georgia Institute of Technology, Atlanta, USA

Date/Time: Nov 11 17:00:00

Location: Online Seminar

Abstract:
Machine learning (ML) has revolutionized artificial intelligence (AI), enabling
breakthroughs across various domains. However, ML??s effectiveness is fundamentally
constrained by the availability and quality of data, posing challenges for applications in many
scientific and engineering fields, where large-scale data collection is impractical or
resource-intensive. To address this challenge, my research focuses on developing hybrid
solutions that integrate scientific knowledge from the target application domains into the ML
models and algorithms. In this talk, I will discuss two such approaches for infusing domain
knowledge into ML methods, illustrated through examples from my research on applying ML in
scientific computing and robotics. First, I will present a direct approach in which numerical
models are structurally embedded into deep neural networks (DNNs) to learn spatiotemporal
physical processes. In the second part, I will illustrate another approach that constrains DNNs
with control-theoretic properties to design controllers for nonlinear systems. Finally, I will
conclude the talk by outlining my future research direction toward building intelligent and
autonomous systems for exploring, monitoring, and operating in dynamic environments.

This is an online seminar and the teams meeting link for this is:
Link

Speaker Bio:
Priyabrata Saha is a Postdoctoral Fellow at the Georgia Institute of Technology. He earned his B.Tech. and M.Tech. degrees in Electronics and Electrical Communication Engineering from the Indian Institute of Technology Kharagpur in 2015, and his Ph.D. in Electrical and Computer Engineering from the Georgia Institute of Technology in 2023. He has published over 20 papers in refereed journals and conferences. His research interests broadly lie in developing machine learning-based solutions for scientific and engineering applications. His current work focuses on building machine learning models for complex physical processes, learning-based optimization and control, and intelligent sensing for autonomous systems.

Host Faculty: R Govindarajan