Heterogeneity-Aware Algorithms for Federated Optimization

Explore federated optimization and heterogeneity-aware algorithms in this 34-minute Google TechTalk presented by Gauri Joshi at the 2022 Workshop on Federated Learning and Analytics. Delve into the objectives and notation of federated optimization, focusing on the FedAvg algorithm. Examine various sources of heterogeneity in federated learning, including data, communication, and computational heterogeneity. Learn from Gauri Joshi, an associate professor at Carnegie Mellon University, recognized for her innovative work in distributed optimization and parallel computing, and recipient of numerous awards including MIT Technology Review's 35 Innovators under 35 and the NSF CAREER Award.

Intro
Federated Optimization: Objectives and Notation
Federated Optimization: The FedAvg Algorithm
Sources of Heterogeneity in Federated Learning
Data Heterogeneity
Communication Heterogeneity
Computational Heterogeneity