Farzad Yousefian Leads Rutgers Research Efforts on Collaborative $7 Million DOE Federated Learning Project
As a Co-PI leading Rutgers research efforts on a three-year collaborative project, Assistant Professor Farzad Yousefian has received $525,000 from a $7 million grant from the Department of Energy’s (DOE) Advanced Scientific Computing Research (ASCR) program.
He is collaborating with three national laboratories – Argonne National Laboratory, Brookhaven National Laboratory, Oak Ridge National Laboratory – and Arizona State University on “Privacy-Preserving Federated Learning for Science: Building Sustainable and Trustworthy Foundation Models.”
“Federated learning (FL) has recently emerged as a collaborative framework for machine learning in applications where the data is distributed among several clients,” Yousefian explains. “This project aims to design, analyze, and numerically implement communication-efficient and privacy-preserving FL algorithms for extreme scale problems, including foundation models.”
The multi-million dollar Argonne-led project is closely related to Yousefian’s recently completed two-year DOE-funded FL project, which designed, analyzed, and numerically validated new algorithms with performance guarantees for addressing hierarchical optimization problems arising in machine learning and game theory.
“These promising results strongly support my research in this newly funded DOE grant,” Yousefian says. “It is truly exciting to be working with mathematical scientists and faculty from national labs and universities on a multi-year funded project that’s well-aligned with my research interest in multi-agent optimization, as well as my professional career goals.
“I also look forward to contributing to the training of my team of graduate and undergraduate students.” His team includes two doctoral students who will be actively involved in all aspects of the projects, as well as undergraduates whom he will mentor in collaboration with Rutgers’ Aresty Research Center.
