Plant a Touch-Me-Not: Train Models Without Anyone Touching Your Data with Flower
summary
Claims use federated learning with Flower to train without moving raw data; evaluation compares PySyft, TensorFlow Federated, and Flower and chooses Flower for framework agnosticism and docs. A demo spins up a containerized app with three clients running three rounds of federated averaging, shows loss decreases and rising accuracy, and publishes the GitHub repo used for the demo and configs for scaling to thousands of clients.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:04Capital One breach privacy concerns
- 4:01Classical ML sensitive data challenges
- 5:27Distributed ML limitations
- 7:11Federated learning model to data
- 8:36Federated learning drawbacks
- 9:38Evaluating FL tools
- 10:43Why Flower stood out
- 11:04How Flower works
- 12:49Live demo setup
- 14:38Configuring Flower server
- 17:12Starting Flower server
- 18:58Creating Flower clients
- 21:08Running federated training
- 23:18Aggregating metrics performance
- 25:08Scaling Flower thousands clients
- 25:53Flower privacy scalability benefits
- 26:58Q&A
related talks
57mKeynote: Federated Learning with Flower: AI's Next Frontier
32mInnovation in the Age of Regulation: Federated Learning with Flower
01h 22mThe future of AI training is federated
18mEnabling Medical AI through Federated Learning
26mMachine Learning without the Data #federatedlearning
31mFederated Machine Learning with Kubernetes
42mLearning Together, Distributed: An Introduction to Federated Learning- 31mFederated Machine Learning with Kubernetes