Sep 2026· IEEE Internet of Things Journal· Vol 13, pp. 39204-39225· 0 citations· 48 references
Abstract
Federated learning (FL) is a distributed machine learning (ML) paradigm that has been widely used to train ML models on massive amounts of data in edge computing (EC) environments. However, FL faces significant challenges from device heterogeneity, edge dynamics, and limited communication resources. To address these challenges, we propose a communication-efficient semi-asynchronous FL (CSFL) framework. First, the work introduces a threshold adaptive gradient compression (TAGC) algorithm, which can reduce redundant communication rounds and accelerate model convergence by appropriately increasing local computation. Second, we propose an adaptive weight adjustment mechanism (AWAM), which employs a staleness-based decay function and, based on varying data distributions, sets different weight coefficients to mitigate the impact of statistical and system heterogeneity. To tackle edge dynamics, a dynamic node selection algorithm based on deep reinforcement learning (DRL) is proposed. This algorithm enables adaptive adjustment of the number of local models participating in global model aggregation according to environmental changes. Finally, we analyze the convergence bound of CSFL theoretically and conduct extensive experiments on classical datasets to demonstrate the effectiveness of our algorithm. Compared with baseline algorithms, the experimental results indicate that CSFL can effectively decrease bandwidth resource consumption and total training time during the training of edge intelligence models across various datasets and data distributions.
Integrating over-the-air computations into the model aggregation process of federated learning (FL) offers a promising solution to mitigate the communication bottleneck in FL model training. In this approach, all the clients modulate their intermediate parameters, such as gradients, onto the same set of orthogonal waveforms and transmit the resulting signals to the edge server simultaneously. Capitalizing on the superposition property of the radio channel, the server can extract an automatically aggregated global gradient from the received radio signal. However, the limited number of orthogonal waveforms imposes a constraint on the dimensionality of transmittable updates, hindering the adoption of more advanced, but high-dimensional models. In light of this challenge, we propose OFLight, a lightweight, yet effective, gradient compression algorithm tailored for OTA-FL systems. Specifically, in each communication round, the edge server constructs a low-rank projection matrix based on the received gradient matrix from the previous round (initialized with an independent and identically distributed standard normal matrix in the first round) and broadcasts it, along with the global model, to all clients in the system. Based on this matrix, every client projects its locally updated gradient matrix into a low-dimensional subspace through a linear operation. The clients upload only their compressed gradients via OTA computations, and the edge server can perform a linear decompression on the received signal, retrieving the original gradient dimension. Moreover, an error feedback mechanism is incorporated to compensate for the approximation error under aggressive compression. We derive analytical expressions for the convergence rate of both convex and non-convex loss functions, quantitatively demonstrating the effect of OFLight on the OTA-FL training efficiency. We also conduct extensive experiments to corroborate the efficacy of the proposed method.
Jiaqi Zhu, H. Yang, Nikolaos Pappas et al.· IEEE Transactions on Wireles...· 0 citations
Federated learning (FL) on heterogeneous edge networks faces a fundamental tension: standard aggregation protocols assume client homogeneity, yet real-world edge deployments span device tiers with $7 \times$ compute and $\mathbf{1 0} \times$ bandwidth disparities. Slow clients become stragglers that stall synchronous rounds, while uniform gradient compression degrades accuracy on bandwidth-constrained devices. This paper presents FedEdge-Adapt, a novel adaptive federated learning framework that jointly addresses device heterogeneity, straggler mitigation, and communication efficiency without sacrificing model quality. FedEdge-Adapt introduces three tightly coupled mechanisms: (1) tier-aware gradient compression that applies device-class-specific sparsification ratios, (2) drift-corrected aggregation that reweights client updates based on staleness and data heterogeneity, and (3) predictive client selection that anticipates dropout-prone devices using a lightweight resource oracle. We evaluate FedEdge-Adapt on a 30-node heterogeneous edge network over 150 communication rounds using the CIFAR-10 dataset under non-IID distributions ($\alpha=0.5$ Dirichlet) and compare against FedAvg, FedProx, and SCAFFOLD baselines. FedEdge-Adapt achieves 85.44% global accuracy, a 6.85 percentage-point improvement over FedAvg, while simultaneously reducing round latency by $\mathbf{6 7. 4 \%}$, communication overhead by $\mathbf{3 4. 0 \%}$, and client dropout rate by $\mathbf{5 4. 0 \%}$. Convergence is reached in 18 rounds versus 31 for FedAvg. Extended experiments across 100+ rounds confirm long-term stability with no late-stage divergence.
Saher Elsayed, Mohamed Ali, Samer Abubaker et al.· Annual International Compute...· 0 citations
The growing smart devices (SDs) in the Industrial Internet of Things (IIoT) generate complex computations that strain the performance and energy of local processing. Mobile Edge Computing (MEC) addresses this by providing nearby computing resources for low-latency offloading. However, achieving efficient computation offloading under massive device concurrency and densely distributed computation offloadings remains a key challenge. To address this, this paper constructs a multi-server MEC system model for IIoT and introduces Mean-Field Game (MFG) theory to model the offloading competition among SDs. This effectively reduces the dimensionality and complexity of multi-agent interactions. A novel Mean-Field Computation Offloading (MFCO) algorithm is proposed, which combines MFG with Rainbow Deep Q-Network under a Multi-Agent Deep Reinforcement Learning framework. By incorporating advanced components such as distributional value estimation, prioritized experience replay, multi-step learning, and dueling architecture, each SD acts as an autonomous agent, optimizing its policy based on local observations and mean-field approximations. Further enhancements include Boltzmann exploration, adaptive learning rates, and a mean Q-network structure, which improve convergence speed and training stability. Extensive simulations on a large-scale IIoT platform (100 SDs, 9 MEC servers) demonstrate that MFCO reduces computation latency and improves long-term rewards while maintaining robust server performance.
Xinmin Cheng, Chengquan Yu, Lu Gao et al.· IEEE Transactions on Green C...· 0 citations
Energy-Aware Adaptive Quantization and Freezing (EA-AQF), a unified framework that co-optimizes communication and computation, is presented, a unified framework that co-optimizes communication and computation and maintains robust convergence in highly heterogeneous tasks.
Timely and accurate defect classification is critical for ensuring product quality and safety in industrial inspection scenarios. The widespread deployment of Internet of Things (IoT) devices equipped with sensing, computing, and communication capabilities has promoted the development of AI-enabled industrial applications. However, conventional AI approaches typically rely on centralized data collection and processing, which become impractical in real-world IoT environments due to growing privacy concerns and constrained device resources. To address these challenges, this paper proposes a communication-efficient adaptive federated learning algorithm for heterogeneous defect classification tasks. The proposed approach jointly accelerates the training process through three mechanisms: (i) adaptive local updates that balance communication and computation overheads; (ii) parameter compression that trades off communication cost against model accuracy; (iii) joint bandwidth and computation-power allocation that optimizes per-round communication and computation time across participating devices. We further analyze the joint effects of these three mechanisms and provide a convergence analysis. Extensive simulations show that the proposed method achieves competitive classification accuracy while reducing single-round training time by up to 70%.
Shuo He, He-Yang Wei, Congxian Bi et al.· Electronics· 0 citations
Federated learning (FL) applications normally employ large deep learning (DL) models, resulting in excessive communication overhead in the deployment of FL over resource-constraint mobile edge networks. To achieve better scalability for DL-based FL, we capitalize on both the asymmetric nature of mobile networks and the distinct effects of partial transmissions on FL training for the global and local models. We propose Fed-DynAmal, an FL framework that decreases the number of parameters transmitted in the uplink (clients-to-server) while concurrently achieving better model performance. The underlying idea is that each selected client sends a partial DL model to the server by omitting several sub-blocks from the trained local model. Crucially, we drop the assumption that transmitted local models can still be used for inference, thereby allowing for greater model variability. At the server, we introduce amalgamation, a process to merge different partial local models into an inference-viable full model. Essentially, amalgamation is a bridge for performing aggregation at the sub-block level. Interestingly, as the key takeaway, communication efficiency versus model performance is not necessarily a trade-off in FL: Our extensive experiments show that Fed-DynAmal can effectively improve communication efficiency while still concurrently achieving higher accuracy and enhanced robustness.
Zihan Chen, H. Yang, Tony Q. S. Quek et al.· IEEE Transactions on Cogniti...· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduAug 27, 2026
A new machine-learning framework aims to improve the success rate of computational protein design while moving away from results that reproduce sequences found in nature.
MIT News · Artificial Intelligence· news.mit.eduAug 3, 2026
