From Theory to Practice: Advances in Real-World Federated Learning
summary
Open-source NVIDIA FLARE provides a production-ready federated learning SDK with new APIs (collaboration API, model streaming, native tensor transfer) and privacy features (differential privacy, homomorphic encryption, confidential computing). A COVID-19 study runs across 20 hospitals using federated averaging and federated preprocessing, yielding a 38% improvement in cross-site generalizability and external test performance of about 95% sensitivity / 88% specificity. Integrates with PyTorch/TensorFlow and supports cross-silo and hierarchical cross-device deployments.
AI-written summary · about
chapters (11)
- 0:00Federated learning at NVIDIA
- 1:26Why federated learning is needed
- 3:58NVIDIA FLARE framework
- 5:26COVID-19 federated learning study
- 10:08Data standardization challenges
- 13:44Healthcare and pharma federations
- 16:22NVIDIA FLARE integration demo
- 21:23Collaboration API for federated workflows
- 26:50Data heterogeneity solutions
- 29:44Privacy efficiency research
- 30:30Open source resources
related talks
30mFederated Learning from simulation to production
31mFederated Learning: So Close and Yet So Out of Reach
32mOpenFL: A Federated Learning Project to Power (and Secure) Your Projects
42mFederated Learning: Unlocking the Values of Discrete Data for Enterprise
27mFederated Learning with Azure Machine Learning NVIDIA FLARE and MONAI
26mOpenFL (Federated Learning) Building better AI models with private data
42mLearning Together, Distributed: An Introduction to Federated Learning
41mFederated Learning: Machine Learning on Decentralized Data