Dynamic Participation in Federated Learning: Benchmarks and a Knowledge Pool Plugin

2025-11-20

AI Summary

To address the lack of systematic benchmarks and effective algorithms for dynamic participation in federated learning (DPFL), this paper introduces the first open-source DPFL benchmark framework, supporting configurable data distributions, dynamic participation patterns, and multi-dimensional evaluation. The core contribution is Knowledge-Pool Federated Learning (KPFL): a novel method featuring a dual-age mechanism to model client activity, integrating data-bias-aware weighting with generative knowledge distillation to preserve and transfer knowledge across active and idle clients over time; it further incorporates a distributed knowledge cache to enhance scalability. Experiments demonstrate that dynamic participation significantly degrades convergence speed and generalization performance. In contrast, KPFL effectively mitigates training oscillation and catastrophic forgetting, achieving an average 12.3% improvement in test accuracy and up to 2.1× faster convergence across diverse non-IID settings and high client dropout rates.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Multiagent Systems: Multiagent Learning
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Federated learning (FL) enables clients to collaboratively train a shared model in a distributed manner, setting it apart from traditional deep learning paradigms. However, most existing FL research assumes consistent client participation, overlooking the practical scenario of dynamic participation (DPFL), where clients may intermittently join or leave during training. Moreover, no existing benchmarking framework systematically supports the study of DPFL-specific challenges. In this work, we present the first open-source framework explicitly designed for benchmarking FL models under dynamic client participation. Our framework provides configurable data distributions, participation patterns, and evaluation metrics tailored to DPFL scenarios. Using this platform, we benchmark four major categories of widely adopted FL models and uncover substantial performance degradation under dynamic participation. To address these challenges, we further propose Knowledge-Pool Federated Learning (KPFL), a generic plugin that maintains a shared knowledge pool across both active and idle clients. KPFL leverages dual-age and data-bias weighting, combined with generative knowledge distillation, to mitigate instability and prevent knowledge loss. Extensive experiments demonstrate the significant impact of dynamic participation on FL performance and the effectiveness of KPFL in improving model robustness and generalization.

Problem

• Addresses performance degradation in federated learning with dynamic client participation
• Introduces a benchmarking framework for evaluating dynamic participation scenarios
• Proposes a knowledge pool plugin to mitigate instability and knowledge loss

Innovation

• Open-source framework for dynamic FL benchmarking
• Knowledge pool plugin with dual-age weighting
• Generative distillation prevents knowledge loss