SPS Webinar: Robust Aggregation for Federated Learning

Date: 10 June 2024
Time: 11:00 AM ET (New York Time)
Presenter(s): Dr. Krishna Pillutla

Original article publicly available for download on the day of the webinar for 48 hours:: Download article


This webinar will present a novel approach to federated learning that endows its aggregation process with greater robustness to potential poisoning of local data or model parameters of participating devices. The proposed approach, Robust Federated Aggregation (RFA), relies on the aggregation of updates using the geometric median, which can be computed efficiently using a Weiszfeld-type algorithm. RFA is agnostic to the level of corruption and aggregates model updates without revealing each device’s individual contribution. The presenter will establish the convergence of the robust federated learning algorithm for the stochastic learning of additive models with least squares and will also offer two variants of RFA: a faster one with one-step robust aggregation, and another one with on-device personalization. He presents experimental results with additive models and deep networks for three tasks in computer vision and natural language processing. The experiments show that RFA is competitive with the classical aggregation when the level of corruption is low, while demonstrating greater robustness under high corruption.


Krishna Pillutla (M ‘21) received the B.Tech. degree from IIT Bombay in Mumbai, India in 2014, the M.S. from Carnegie Mellon University in Pittsburg, PA, USA in 2015, and the Ph.D. from University of Washington in Seattle, WA, USA in 2022, in computer science and engineering.

He is an assistant professor at IIT Madras in Chennai, India. Previously, he was a visiting researcher (postdoc) at Google Research, USA. His research interests include privacy-preserving, federated, and robust machine learning.

Dr. Pillutla’s awards and honors include the NeurIPS Outstanding Paper Award (2021), the J.P. Morgan Ph.D. Fellowship (2019-20), and the Anne Dinning - Michael Wolf Endowed Regental Fellowship (2016-17).