Chapter 9: Federated Learning Overview

Learning Objectives

• Formulate the federated-learning problem with $n$ users, a central server, and per-user local datasets
• State and apply the FedAvg algorithm and understand its convergence properties on i.i.d. and non-i.i.d. data
• Quantify the per-round communication cost of FL and identify the dominant bottlenecks
• Survey communication-efficient techniques: quantization, top-$K$ sparsification, structured updates
• Articulate the privacy concern: individual gradients leak information about local data (gradient inversion)
• Motivate why Chapters 10–12 require explicit privacy-preserving protocols on top of FL