2026· IEEE Transactions on Cognitive Communications and Networking· Vol 12, pp. 10514-10529· 0 citations· 45 references
Computer Science
Abstract
Semi-Asynchronous Federated Learning (SAFL) takes advantage of both synchronous and asynchronous FLs to train models. However, existing works in semi-asynchronous FLs fail to fully account for heterogeneities in both data and devices. To address these issues, we first propose a Clustered SAFL (CSAFL) framework and theoretically analyze its convergence loss. Then, a convergence loss minimization problem is formulated under the considerations of heterogeneities in device resources, fairness of cluster selection, and data heterogeneity. To address this complex problem due to nonlinearities and multi-dimensional decision variables, we first design a device clustering algorithm based on both devices’ model parameter differences and gradient directions between local and global models. Then, the original loss minimization problem is transformed into inter- and intra-cluster selection problems. For the inter-cluster selection problem, to ensure fairness, we propose a reinforcement learning-driven Lyapunov approach to fairly select clusters, where reinforcement learning (RL) is used to supervise the cluster selection results by the Lyapunov method. For the intra-cluster selection, we convert it into a constrained multi-armed bandit (MAB) problem in order to let devices within a cluster submit models synchronously. Then, a two-stage Upper Confidence Bound (UCB) scheme is proposed to obtain device selection results. Extensive numerical results with baselines show that our approach achieves up to 25% higher accuracy.
A new hybrid decomposition approach for optimizing multi-model federated learning (MMFL) in edge computing environments that enhances the speed of convergence, contention of resource, and real-time performance, which makes it especially appropriate to apply it to the real-world, e.g., smart healthcare, autonomous vehicles, and IoT systems.
M. Sophiya, Sugantha Grace, N. Brintha· Knowledge and Information Sy...· 0 citations
Heterogeneous federated learning leads to system and data differentials that cause stragglers to either be a bottleneck to synchronous optimization or create representation bias in asynchronous contexts. Although current approaches deal with staleness or buffering independently, their approach does not ensure fast clients do not take over the global model. The proposed framework Straggler-Aware Asynchronous Federated Learning (SAFL), that re-defines the stragglers as structured subjects rather than outliers. SAFL employs temporal exponentially weighted moving average signature of client costs and costs model updates by clustering costs in time-constrained per-cluster buffers. An innovative fairness-sensitive aggregation scheme then balances the participation through frequency compensation and damping on staleness. The results of the experiment indicate that SAFL achieves a 75% accuracy in 620 seconds, 27% higher than the state-of-the-art Federated Asynchronous Mobile Update (FedASMU) and increases the fairness index by 0.52 to 0.87. SAFL has a scalable, fair approach to the regulation of heterogeneous clusters, which means they can be used to ensure almost equal contribution in regulated settings such as financial and healthcare analytics.
S. Babalola· 2026 7th International Confe...· 0 citations
In this article, a novel homotopic reinforcement learning (RL) framework is proposed to address the distributed cluster consensus problem in continuous-time multiagent systems (MASs). For the first time, the investigated issue is formulated as a zero-sum differential game using a proposed minmax game policy, in which a local regulation error is introduced to characterize the deviation of each agent from its assigned leader. Based on this formulation, a set of group game algebraic Riccati equations is derived to obtain the optimal control law. To overcome reliance on known system models, these equations are solved using a data-driven homotopic policy-iteration scheme that leverages online state and input information. In contrast to conventional learning schemes, the proposed approach embeds a homotopic process that relaxes the requirement for an admissible initial policy. Rigorous stability and convergence analyses are provided, and the effectiveness of the proposed method is further demonstrated through theoretical analysis and numerical simulations.
Qing Yang, Jing Wang, Jinde Cao et al.· IEEE Transactions on Cyberne...· 0 citations
This paper develops a hybrid offline-online multi-agent reinforcement learning framework based on decision transformers. The policy is first pretrained offline via supervised sequence modeling of trajectories generated by existing policies, providing a safe and sample-efficient initialization. It is then fine-tuned online using a hybrid objective that incorporates critic-guided gradients, enabling performance improvements beyond the offline policy. To facilitate stable offline-to-online transfer and effective multi-agent coordination, the framework incorporates return-weighted sampling, a critic conditioned on neighbors'actions, and neighborhood-correlated exploration. The approach is fully distributed: both training and execution rely only on local observations and limited information exchange among neighboring agents. Evaluations with dynamic traffic arrivals in two settings: (i) joint scheduling and power allocation and (ii) coordinated beamforming, show that the proposed method achieves quality-of-service (QoS) performance comparable to centralized methods. Moreover, when pretrained on lower-quality datasets, online fine-tuning is also observed to surpass the initial offline policy. These results demonstrate a promising learning-based alternative for wireless resource management.
Yiming Zhang, Kun Yang, Cong Shen et al.· 0 citations
With the rapid development of the Internet of Things (IoT) and edge computing, Federated Learning (FL) has emerged as a promising distributed framework capable of effectively leveraging distributed devices for machine learning tasks while preserving data privacy. However, in practical scenarios characterized by significant device heterogeneity and communication delays, traditional synchronous FL often suffers from inefficiency due to the straggler effect. Although Asynchronous Federated Learning (AFL) has been proposed to mitigate this, it still faces challenges regarding model staleness and training bias, which hinder convergence performance. To address these issues, this paper presents an asynchronous framework named FedQS. First, FedQS employs a multi-dimensional staleness evaluation mechanism that dynamically assesses updates by combining the similarity between local and global models with client latency metrics. Second, to resolve the coupling between training bias and model staleness, we implement a decoupling solution via a queue scheduling algorithm. This algorithm reprocesses high-staleness model parameters on selected faster helper clients using only the helpers’ local private data while preserving the provenance of the original uploading clients, thereby aiming to reduce stale-update effects in aggregation. Finally, during the aggregation phase, the framework recalibrates aggregation weights based on client contributions to reduce training bias and improve global model accuracy. Comparative experiments on Fashion-MNIST and CIFAR-10 datasets demonstrate that FedQS outperforms four evaluated representative baselines—FedAsync, FedBuff, KAFL, and CAFL—achieving an average accuracy improvement of 8.62%.
Jia-Hui Zhou, Fang Li, Tian-Yu Shi et al.· Journal of Cloud Computing· 0 citations
Over-the-air FL with EH MDs under heterogeneous data distributions under heterogeneous data distributions is studied, and the proposed unified framework improves fairness or personalization, depending on the operating mode, while reducing communication overhead.
F. Bagci, Busra Tegin, Mohammad Kazemi et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
