Jun 2026· 2026 2nd International Conference on Federated Learning and Intelligent Computing Systems (FLICS)· pp. 100-106· 0 citations· 31 references
Computer Science
TL;DR
Experimental results show that FedOPAL not only significantly outperforms the original analytical methods on several benchmarks, but also achieves accuracy comparable to state-of-the-art iterative methods while maintaining zero server-side training costs, providing a new engineering paradigm for efficient collaboration of large models on the edge.
Abstract
With the widespread deployment of basic models in edge intelligence, communication bandwidth has become a core bottleneck restricting the scalability of federated learning. Although one-shot federated learning alleviates this problem by minimizing communication rounds, existing iterative fine-tuning or knowledge distillation methods still face challenges such as high server-side computational costs and hyperparameter sensitivity. Analytical federated learning achieves efficient gradientfree aggregation using least-squares closed-form solutions, but in environments with non-independent and identically distributed data, its static feature assumptions fail, leading to feature manifold misalignment and severely impairing model performance. To address this contradiction, this paper proposes the FedOPAL framework. This framework adapts the visual prompts as feature rectifiers, actively correcting the feature distribution of heterogeneous data to a linearly separable space by applying local proximal constraints, thereby satisfying the theoretical assumptions of analytical federated learning. Experimental results show that FedOPAL not only significantly outperforms the original analytical methods on several benchmarks, but also achieves accuracy comparable to state-of-the-art iterative methods while maintaining zero server-side training costs, providing a new engineering paradigm for efficient collaboration of large models on the edge.
Federated fine-tuning is bottlenecked by communication: FedAvg and pseudo-gradient schemes transmit a payload that scales with the model, and gradient compression shrinks it by only a constant factor. We take a different lever. Mapping networks generate a network's weights from a small trainable latent through a frozen affine projection; because the map is shared and affine, averaging latents is exactly averaging the generated weights. We turn this into a practical low-bandwidth federated channel with two changes: a low-rank, seed-regenerable factorisation of the projection (cutting generator memory from ~80 GB to ~10 MB), and a delta formulation $\theta = \theta^{\mathrm{pre}} + U V^{\top} z$ that learns an additive correction around a shared centrally-pretrained base -- federated fine-tuning, which is what makes the method work at scale. A frozen orthogonal classifier head further removes the head from the payload while improving accuracy. On CIFAR-100 with ResNet-18+GroupNorm, our method (FLITE, Federated Low-rank Iterative Training Engine) communicates 1,280 floats (~5 KB) per client per round -- an 8718x reduction -- and reaches 74.67%, within ~0.5 pp of full-weight FedAvg. The averaging identity holds to floating-point precision ($6 \times 10^{-8}$); the method sits one to two orders of magnitude below PowerSGD and top-k on the bandwidth-accuracy Pareto; it matches or exceeds full-weight FedAvg under strong non-IID skew. int4 latents reach 648 bytes per round at unchanged accuracy, whereas int4 full-weight FedAvg collapses to chance.
Experimental results on three datasets demonstrate that FedLIM achieves superior global model accuracy compared to existing one-shot FL methods, particularly in highly heterogeneous environments.
Yexin Dou, Haijiang Wang, Jian Wan et al.· IEEE Internet of Things Jour...· 0 citations
This work proposes RAVEL-FCL, a generative replay-based framework for federated continual learning that integrates an improved generative model based on Rebooting ACGAN with multi-level feature alignment to ensure consistency and employs Elastic Variational Continual Learning on the server to probabilistically regularize the global model and preserve past knowledge.
Yurui Zhou, Jia Hu, Geyong Min et al.· ACM Transactions on Autonomo...· 0 citations
In industrial internet of things (IIoT) deployments, condition monitoring data is rarely complete. Labeled samples are scarce, sensors drop out, signals are corrupted by noise, and what reaches the edge is often a sparse, distorted fragment of the ideal dataset. Federated learning preserves privacy, yet its standard formulations assume clients can train locally on adequate labeled data-an assumption that seldom holds on the factory floor. This paper proposes prompt transfer, a distinct federated learning paradigm that replaces conventional gradient or prototype exchange with the transfer of small, curated sets of labeled examples (prompts) as the sole medium of knowledge flow. Unlike federated distillation, which shares model outputs, or meta-learning, which requires local gradient adaptation, prompt transfer leverages the in-context learning capability of a frozen tabular foundation model to achieve zero-gradient, parameter-free knowledge transfer across clients. Experiments on the Southeast University bearing dataset show that under 5-shot settings, our method FedICL achieves 94.0% accuracy, in contrast to FedAvg’s random-guessing performance of 20.1% while requiring zero gradient computation and reducing communication rounds by 95% and data transfer by 87%. The approach remains effective under non-IID distributions ( 53.9% vs 20% random). This work directly addresses the special issue themes of incomplete data and federated learning, offering a practical solution for label-scarce, resource-constrained IIoT scenarios.
Haoyu Gu, Diming Zhang, Shichao Yi et al.· Measurement science and tech...· 0 citations
A domain-sensitive federated pruning framework that preserves domain-invariant structures while retaining domain-specific representations and a structure-aware aggregation algorithm that fuses heterogeneous personalized architectures into a domain-generalized global model is proposed.
Chenchen Lin, Wenhao Yuan, Zhengji Xu et al.· 0 citations
FeDiSyn is proposed, a unified framework that holistically considers the interplay between pre-training and FedFT to minimize the overall LVM training time and introduces a scaling law for FedFT pre-training to determine the optimal number of synthetic images, balancing pre-training benefit against generation/pre-training cost.
Qianpiao Ma, Xiaozhu Song, Junlong Zhou et al.· 0 citations