PTOPOFL: Topology-Guided Personalised Federated Learning with Persistent-Homology Client Descriptors

Name: PTOPOFL: Topology-Guided Personalised Federated Learning with Persistent-Homology Client Descriptors
Uploaded: 2026-10-21T01:06:13.920314+00:00
Duration: 606 s
Description: PTOPOFL replaces sensitive model updates with compact topological descriptors derived from persistent homology to solve client heterogeneity and privacy in federated learning. The method clusters clients based on these descriptors, allowing for personalized model training within each cluster. This approach provides a provable reduction in reconstruction risk compared to standard gradient sharing.

PTOPOFL replaces sensitive model updates with compact topological descriptors derived from persistent homology to solve client heterogeneity and privacy in federated learning. The method clusters clients based on these descriptors, allowing for personalized model training within each cluster. This approach provides a provable reduction in reconstruction risk compared to standard gradient sharing.

Open the full episode