Federated Learning
Train models on distributed data without centralizing it — privacy-preserving ML. Federated learning enables training on data that cannot leave the source.
Level: Advanced · Category: Machine Learning · Estimated time: 6 hours
Prerequisites
- Machine Learning Basics
- Neural Networks Deep Dive
Lessons
- Federated Learning Fundamentals — Motivation, privacy, and the federated setting.
- Federated Averaging — FedAvg algorithm, client updates, and server aggregation.
- Secure Aggregation — Cryptographic protocols for privacy-preserving aggregation.
- Differential Privacy in FL — Adding noise, privacy budgets, and epsilon-delta.
- Non-IID Data Challenges — Personalization, client drift, and solutions.
- Implementing with Flower — Client setup, server, and running federated training.
Topics covered
federated-learning, privacy, distributed, differential-privacy