Learning Together, Distributed: An Introduction to Federated Learning
summary
Overview of federated learning fundamentals with concrete results: models move instead of raw data, used by Gboard and Apple for on-device language/speech models with differential privacy protections; demonstration uses the Python framework Flower to run FedAvg aggregation on an MNIST classifier split across 20 clients (scalable to 500), shows loss decreasing in rounds, and runs client/server via Docker. Covers practical limits: model poisoning, client dropouts, non-IID fairness and asynchronous strategies.
AI-written summary · about
chapters (10)
- 0:00Introduction and motivation
- 3:55Classic machine learning limits
- 7:07Federated learning basics
- 10:39Real-world federated learning use cases
- 13:32Federated learning frameworks
- 16:46Flower framework overview
- 21:40Flower demo setup
- 26:33Flower demo with Docker
- 29:41Federated vs centralized learning
- 33:12Learning resources and future
related talks
57mKeynote: Federated Learning with Flower: AI's Next Frontier
41mFederated Learning: Machine Learning on Decentralized Data
34mDecentralized Machine Learning
32mInnovation in the Age of Regulation: Federated Learning with Flower
26mMachine Learning without the Data #federatedlearning
31mFrom Theory to Practice: Advances in Real-World Federated Learning
34mTraining without sharing: Is Federated Learning the answer to data privacy in AI?
31mFederated Machine Learning with Kubernetes