FedLess: Secure and Scalable Serverless Federated Learning
Talkat Open Source Summit Europe 2023 · on Tuesday, 2023-09-19· at 12:20· Room 0D-2-0D-3 (Floor 0), Abandoibarra Etorb., 4, 48011 Bilbao, Bizkaia, Spain·
summary
FedLess is an open‑source, cloud‑agnostic serverless FL framework that runs client functions on AWS Lambda, Google Cloud Functions and Azure Functions, uses MongoDB as parameter server, is written in Python 3 with TensorFlow/Keras support, and applies local differential privacy (Gaussian noise) by default. It implements a semi‑asynchronous FedScan strategy using DBSCAN clustering and staleness‑aware aggregation, demonstrates up to 200 concurrent clients, speeds convergence enough to cut training time ~8% and cost ~20%, and raises final accuracy ~2% with 70 simulated stragglers.
AI-written summary · about
chapters (9)
- 0:00FedLess introduction
- 1:03Cloud computing evolution
- 4:52Federated Learning basics
- 7:03Serverless + Federated Learning
- 8:11FedLess architecture
- 13:25Addressing stragglers in Federated Learning
- 20:55Results and cost analysis
- 24:50Personalized models
- 28:03Q&A
related talks
31mFedLess: Secure and Scalable Serverless Federated Learning
42mLearning Together, Distributed: An Introduction to Federated Learning
42mFederated Learning: Unlocking the Values of Discrete Data for Enterprise
31mFrom Theory to Practice: Advances in Real-World Federated Learning
41mFederated Learning: Machine Learning on Decentralized Data
30mFederated Learning from simulation to production
45mManaged federated learning on AWS: A case study for healthcare
31mFederated Machine Learning with Kubernetes