Trustworthy, Efficient, and Robust Distributed Systems
Federated Learning (FL) has emerged as a crucial paradigm for training machine learning (ML) models on decentralized edge devices while addressing data privacy. In FL, models are collaboratively trained across multiple clients without sharing raw data, addressing privacy concerns while enabling collaborative ML. In this talk, I will delve into the myriad practical challenges encountered in the implementation of FL, and present my research contributions aimed at overcoming these obstacles.
Drawing inspiration from diverse fields like information theory, data privacy, and optimization, my research works offer innovative solutions to advance the state-of-the-art in FL. These contributions are poised to make privacy-preserving large-scale ML a practical reality. Furthermore, my talk will encompass my recent research focusing on ‘exact machine unlearning’ for federated clustering (FC), addressing privacy concerns even ‘after’ model training. This novel perspective has gained significance in the wake of emerging data privacy laws and the ‘right to be forgotten’, particularly for sensitive data like medical and genomic records.
Saurav Prakash is a postdoctoral fellow at the Institute for Genomic Biology at the University of Illinois Urbana-Champaign (UIUC), where he is working in genomic security and privacy with Prof. Carl Gunter and Prof. Olgica Milenkovic. He received the B. Tech. degree in Electrical Engineering from IIT Kanpur in 2016, and a Ph.D. degree in Electrical and Computer Engineering from the University of Southern California (USC) in 2022, where he was advised by Prof. Salman Avestimehr. His research interests include distributed and privacy-preserving machine learning, especially federated learning. Among his accolades, he is one of the recipients of the Qualcomm Innovation Fellowship in 2021. He also received the Annenberg Graduate Fellowship in 2016 and was one of the Viterbi-India Fellows in summer 2015.