Demystifying Decentralized Fair Models for Federated Machine Learning
summary
Claims a practical pipeline that blends PySyft-based federated training with additive secret sharing and differential privacy (Laplace/Gaussian) to produce cluster-aware models for IoT and healthcare. Presents a three-layer client–edge–server architecture, variational clustering on an IoT male/female dataset, a tunable fairness penalty lambda, and privacy budget epsilon; reports comparable accuracy to conventional FL while improving proportional representation and robustness, and points to a GitHub demo and appendix for reproducible code.
AI-written summary · about
chapters (16)
- 0:00Intro
- 1:49Data and Privacy
- 3:54Centralized Machine Learning (ML) - Typical
- 5:15Introduction to Federated Learning
- 8:00Levels of Federated Learning
- 9:29Federated Learning Basic Linear Model Code
- 10:26Additive Secret Sharing & Fixed Precision Encoding
- 11:35Differential Privacy & Federated Learning
- 12:41Federated Learning - Benefits and Challenges
- 16:14Federated Learning with Clustering
- 21:01Federated Learning Complete Evolution
- 22:39Coreset and Variation Framework for Fair Clustering
- 24:00Clustered Federated Learning in Distributed DP Systems
- 26:32Architecture - Federated Learning in a nutshell
- 28:06Conclusion/Summary to complete a Federated Model
- 29:42Variational Clustering Fair K-Means
related talks
26mMachine Learning without the Data #federatedlearning
27mCreate Value and Insight Through Privacy-preserving Data Collaboration [...]
29mBalancing Thin Line Between Data Intelligence & Privacy
31mFederated Machine Learning with Kubernetes
28mA Privacy Enhancing Federated Learning Framework for Multi-Party Distributed ..
34mDecentralized Machine Learning
31mScaling privacy embedded distributed ML systems
42mLearning Together, Distributed: An Introduction to Federated Learning