Hierarchical federated learning (HFL) emerges as a promising solution for distributed systems in real-world IoT, compensating for the limitations of client and training data through multilevel aggregation. However, severe system heterogeneity, such as data heterogeneity, differentiated communication and computing power, can cause model drift and negatively impact system performance, posing challenges to existing HFL frameworks. To address this issue, we propose a mitigating edge heterogeneity method for HFL (MEH-Fed) in this paper. First, we design a new HFL framework, where the key idea is to introduce an initial update of the edge server side to reduce the impact of heterogeneity on system performance. Edge servers use synthetic datasets to guide the update direction of client models, facilitating model training and accelerating convergence. Then, we conduct a convergence analysis of the proposed method, establish a theoretical upper bound, and analyze the effects of key parameters. Finally, we experimentally evaluate the performance of the proposed MEH-Fed in terms of system performance and convergence rate. Experimental results validate that MEH-Fed achieves optimal performance in both accuracy and efficiency, establishing a new paradigm in HFL architecture design for heterogeneous IoT environments.
D. Wu, Changxin Lu, Zhi-Yuan Zhu et al.· IEEE Transactions on Mobile...· 0 citations
Remote sensing change detection (RSCD) remains vulnerable to pseudo-changes caused by seasonal, illumination, and atmospheric discrepancies between bi-temporal images. Existing deep models often ignore this temporal distribution gap or align entangled features directly, which may corrupt change-relevant semantics and cause negative transfer. To address this challenge, this article proposes the disentangled adversarial alignment network (DAANet), establishing a novel “Disentangle to Align” paradigm and introducing a domain adaptation (DA)-inspired temporal distribution alignment framework for time-invariant feature learning (IFL) in change detection. DAANet first uses a content-style gate (CSG) to separate content-preserving features from style-biased residuals, and then applies adversarial alignment only to the residual branch. This targeted alignment encourages the backbone to learn time-invariant representations while preserving semantic cues for real changes. A dual-dimensional dynamic balancing strategy further stabilizes the adversarial optimization. Extensive experiments on WHU-CD, LEVIR-CD, SYSU-CD, and MSRSCD, together with backbone-universality analyses, demonstrate the effectiveness and robustness of DAANet under complex imaging conditions.
