FedDAAW: Dynamic Client Selection and Accuracy- Adaptive Aggregation for Cross-Task Federated Learning under Heterogeneity
Abstract
Abstract Federated learning enables collaborative model training across distributed clients while preserving data privacy. However, both statistical and system heterogeneity among clients severely degrade model performance. Existing methods mostly address only one aspect of heterogeneity and are validated primarily on classification tasks, with limited evidence of cross-task generalization.To address these issues, this paper proposes FedDAAW, a lightweight framework for heterogeneous federated learning. It introduces a multi-factor dynamic client selection mechanism and an accuracy-adaptive aggregation strategy, which fully leverages high-quality clients without requiring gradient sharing or auxiliary data.Experiments are conducted on image classification (MNIST, CIFAR-10), text sentiment analysis (IMDB), and regression (California housing) datasets. Under highly non-IID settings, FedDAAW consistently outperforms the baseline FedAvg across all tasks, achieving test accuracies of 98.80%, 31.05%, and 83.86% on MNIST, CIFAR-10, and IMDB, respectively, and reducing the mean squared error to 0.2918 on the California housing dataset. Furthermore, the proposed method shows significant advantages in convergence efficiency, reaching predefined performance targets in fewer communication rounds and thus lowering communication overhead.FedDAAW is privacy-preserving, lightweight, and easy to deploy. It effectively improves the overall performance of heterogeneous federated learning…
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.