Abstract:
Machine learning (ML) is increasingly playing a pivotal role in the sciences. Yet, a number of open questions remain on the best learning strategies to maximize the utility of ML, particularly in limited data scenarios. In this talk, I will discuss some of our methods to address these challenges, with a central theme that math and physics tricks can be powerful when combined with neural networks. I will show examples on how such “tricks” can allow us to: change our basis of learning via spectral methods to solve challenging fluid dynamics and transport PDE problems, combine neural networks with differentiating through physics-based numerical solvers for more precise enforcement of conservation laws and more accurate spatiotemporal modeling, and greatly improve the efficiency of equivariant neural networks for the E(3) group, leading to better materials property predictions. In all of these examples, we show that such learning strategies enhance our capabilities to handle complex scientific problems, bridging gaps between theoretical understanding and practical application.
Bio:
Aditi Krishnapriyan is an Assistant Professor at UC Berkeley where she is a member of Berkeley AI Research (BAIR), the AI+Science group in Electrical Engineering and Computer Sciences (EECS), and the theory group in Chemical Engineering. Her research interests include physics-inspired machine learning methods; geometric deep learning; inverse problems; and the development of machine learning methods informed by physical sciences applications including molecular dynamics and fluid mechanics. Previously, she was a Luis W. Alvarez Fellow in Computing Sciences at LBNL and received a PhD from Stanford University while supported by the DOE Computational Science Graduate Fellowship.
Location:
Microsoft Teams meeting
Join on your computer, mobile app or room device.
Click here to join the meeting
Meeting ID: 225 924 574 733
Passcode: nNRsX3
Download Teams | Join on the web
Or call in (audio only)
+1 630-556-7958,50836801# United States, Big Rock
Phone Conference ID: 508 368 01#