Resilient federated learning framework
Abstract
Federated Learning (FL) offers a promising paradigm for privacy-preserving distributed Machine Learning (ML), yet its real-world applicability is often hindered by the need for robust resilience and adaptability in dynamic network environments, where node failures and changing connectivity are common. This dissertation designs, implements, and evaluates FlexFL, a novel Resilient Federated Learning Framework focusing on modularity and fault tolerance to address these challenges. FlexFL features a highly modular architecture, allowing flexible integration of ML backends, FL algorithms, and communication protocols, and incorporates resilience mechanisms such as dynamic worker pool management and task rescheduling. The framework’s performance and robustness were rigorously evaluated across diverse scenarios, including simulated worker failures on large-scale heterogeneous environments using the UNSW-NB15 and ToNIoT datasets. Notably, in a demanding scenario involving 40 workers with a 1% probabilistic failure rate, FlexFL demonstrated successful training continuity and model convergence for both datasets. These results validate FlexFL’s capability to effectively manage dynamic worker participation and tolerate substantial node failures, positioning it as a robust and scalable solution for real-world FL deployments in challenging edge computing environments.
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.