Suri Vaikuntanathan (UChicago): AI+ Science Schmidt Fellows Speaker Series

Organized by the University of Chicago’s Eric and Wendy Schmidt AI in Science Postdoctoral Fellowship Program.

Agenda
4:00pm – 4:45pm:  Presentation
4:45pm – 5:00pm:  Q&A
5:00pm – 5:30pm: Reception

Meeting location
William Eckhardt Research Center. Room 401
5640 S Ellis Avenue, Chicago, IL 60637
Map It

Abstract: Diffusion models demonstrate robust generative capabilities by approximating dataset distributions and synthesizing data through sampling. In this talk I will discuss our recent work investigating how incorporating temporally correlated noise sources—inspired by active matter physics—affects generative performance during the forward destruction process. We derive the necessary cost function for training models with exponentially correlated noise in the diffusion process and evaluate this approach across diverse datasets: multi-scale 2D distributions, high-dimensional Ising model spin configurations at varying temperatures, and molecular microscopic degrees of freedom. Our results indicate that this active noise approach can enhance the reverse diffusion sampling process compared to standard passive models. Through analytical calculations, we propose that this improvement stems from accelerated speciation time during sampling and more efficient resolution of high-frequency data modes. This physically-motivated modification to score-based diffusion models offers a novel optimization strategy that may reduce both training and inference times across application.

Bio: Suri Vaikuntanathan is a Professor in Chemistry at the University of Chicago. He completed his B.Tech in Biotechnology from the Indian Institute of Technology Madras in 2006, followed by a Ph.D. in Chemical Physics from the University of Maryland, College Park in 2011. After completing his postdoctoral research at the University of California, Berkeley in 2014, he established his independent research career. His group develops theoretical and simulation methods to study far-from-equilibrium systems, investigating principles of assembly and pattern formation under non-equilibrium conditions. They also explore the relationship between energy dissipation and information processing in biological circuits, examining how biological systems maintain sensitivity and robustness in noisy environments. His honors include ACS Early Career Award in Theoretical Chemistry (2023), the Camille Dreyfus Teacher-Scholar Award (2020), an NSF CAREER Award (2018), and an Alfred P. Sloan Fellowship (2017).

Parking
Campus North Parking
5505 S Ellis Ave
Map It