Adaptive Federated Representation Learning for Nonstationary Edge Intelligence
DOI:
https://doi.org/10.36520/joofin.v1.i1.3Keywords:
federated learningAbstract
Federated learning at the edge is frequently evaluated under static client distributions, although real deployments exhibit non-independent and non-identically distributed data together with temporal concept drift. This paper proposes NeuroDrift-FL, a drift-aware federated representation-learning framework that couples compact latent prototypes with adaptive client weighting. Each client estimates feature-distribution movement from consecutive local windows, optimizes a representation compactness objective, and transmits model parameters together with a scalar drift score. The server increases the contribution of informative drifted clients while preserving global stability through bounded weighting. A reproducible controlled benchmark with nine heterogeneous clients, three classes, and abrupt-to-gradual feature drift was used to isolate the mechanism. Across 22 communication rounds, NeuroDrift-FL produced higher post-drift macro-F1 than FedAvg and FedProx and recovered more rapidly after the distribution change. The results indicate that explicitly linking representation stability and drift-aware aggregation can improve adaptation without exchanging raw client data. The study is a computational proof-of-concept; deployment-scale validation on public edge benchmarks is identified as the next step.
Downloads
References
[1] M. Amadeo, C. Campolo, G. Ruggeri, and A. Molinaro, “Improving communication performance of federated learning: A networking perspective,” Computer Networks, 2025, doi: 10.1016/j.comnet.2025.111353.
[2] H. Peng, C. Wu, and Y. Xiao, “FD-IDS: Federated learning with knowledge distillation for intrusion detection in non-IID IoT environments,” Sensors, vol. 25, no. 14, art. 4309, 2025, doi: 10.3390/s25144309.
[3] N. Romandini, A. Mora, C. Mazzocca, R. Montanari, and P. Bellavista, “Federated unlearning: A survey on methods, design guidelines, and evaluation metrics,” IEEE Trans. Neural Netw. Learn. Syst., vol. 36, no. 7, pp. 11697–11717, 2025, doi: 10.1109/TNNLS.2024.3478334.
[4] Y. Wang, H. Hu, and W. Du, “Communication-efficient personalized federated learning with evolutionary strategies and cluster-aware knowledge transfer,” Knowledge-Based Systems, 2025, doi: 10.1016/j.knosys.2025.114246.
[5] R. Teixeira, G. Baldoni, M. Antunes, D. Gomes, and R. L. Aguiar, “Leveraging decentralized communication for privacy-preserving federated learning in 6G networks,” Computer Communications, vol. 233, art. 108072, 2025, doi: 10.1016/j.comcom.2025.108072.
[6] M. Ameen et al., “Addressing unreliable local models in federated learning through unlearning,” Neural Networks, vol. 180, art. 106688, 2024, doi: 10.1016/j.neunet.2024.106688.
[7] A. W. Mamond, M. Kundroo, and T. Kim, “Autoencoder-based decentralized federated learning for efficient communication,” Computer Networks, 2025, doi: 10.1016/j.comnet.2025.111676.
[8] X. Di, X. Fan, L. Chen, M. Li, and M. Zhang, “Communication-privacy-accuracy trade-offs in federated learning for non-IID data with shuffle model,” Knowledge-Based Systems, vol. 324, art. 113872, 2025, doi: 10.1016/j.knosys.2025.113872.
[9] H.-G. Joo, S. Hong, and D.-J. Shin, “FedLSC: Improving communication efficiency and robustness in federated learning with stragglers and adversaries,” IEEE Trans. Neural Netw. Learn. Syst., vol. 36, no. 11, pp. 19805–19819, 2025, doi: 10.1109/TNNLS.2025.3590015.
[10] “Reinforcement-learning-based layer-wise aggregation for personalized federated learning,” IEEE Internet Things J., vol. 12, no. 7, pp. 8614–8625, 2025, doi: 10.1109/JIOT.2024.3502245.
[11] A. Zhang et al., “Decentralized federated learning towards communication efficiency, robustness, and personalization,” ACM Trans. Sensor Netw., vol. 21, no. 3, 2025, doi: 10.1145/3730587.
[12] “A communication-efficient federated learning approach via dynamic mutual distillation for image recognition,” Applied Soft Computing, vol. 180, art. 113286, 2025, doi: 10.1016/j.asoc.2025.113286.
[13] Y. Li, X. Wang, H. Li, P. K. Donta, M. Huang, and S. Dustdar, “Communication-efficient federated learning for heterogeneous clients,” ACM Trans. Internet Technol., vol. 25, no. 2, 2025, doi: 10.1145/3716870.
[14] J. Yun, Y. Oh, Y. S. Jeon, and H. V. Poor, “Communication-efficient federated learning over capacity-limited wireless networks,” IEEE Trans. Cogn. Commun. Netw., vol. 11, no. 1, pp. 621–637, 2025, doi: 10.1109/TCCN.2024.3419039.
[15] “Robust multimodal federated learning for non-IID multimodal data with incompleteness,” Future Generation Computer Systems, 2025, doi: 10.1016/j.future.2025.107948.
[16] Q. Zhou et al., “FedPLC: Federated learning with dynamic cluster adaptation for concept drift on non-IID data,” Sensors, vol. 26, no. 1, art. 283, 2026, doi: 10.3390/s26010283.
[17] X. Chen et al., “Virtual concept drift detection and adaptation in federated data stream learning,” Int. J. Data Sci. Anal., vol. 21, art. 46, 2026, doi: 10.1007/s41060-025-00953-2.
[18] K. Sudhakar et al., “FedDriftGuard adaptive federated learning with differential privacy for concept drift in edge environments,” Scientific Reports, vol. 16, art. 20326, 2026, doi: 10.1038/s41598-026-51535-6.
[19] Q. Wang, A. Liu, C. Feng, Y. Xu, J. Zhong, and V. S. Sheng, “A novel federated learning-based framework defending server-level privacy leakage,” Expert Systems with Applications, vol. 283, art. 127617, 2025, doi: 10.1016/j.eswa.2025.127617.
[20] “TPFL: Privacy-preserving personalized federated learning mitigates model poisoning attacks,” Information Sciences, vol. 702, art. 121901, 2025, doi: 10.1016/j.ins.2025.121901.
[21] “Federated regressive learning: Adaptive weight updates through statistical information of clients,” Applied Soft Computing, vol. 166, art. 112043, 2024, doi: 10.1016/j.asoc.2024.112043.
[22] S. Kalloori and A. Srivastava, “Towards cross-silo federated learning for corporate organizations,” Knowledge-Based Systems, vol. 289, art. 111501, 2024, doi: 10.1016/j.knosys.2024.111501.
[23] “An adaptive asynchronous federated learning framework for heterogeneous Internet of Things,” Information Sciences, vol. 689, art. 121458, 2025, doi: 10.1016/j.ins.2024.121458.
[24] L. Yi et al., “pFedKT: Personalized federated learning with dual knowledge transfer,” Knowledge-Based Systems, vol. 292, art. 111633, 2024, doi: 10.1016/j.knosys.2024.111633.
[25] K. Khan, W. Elibox, T. D. Ramlochan, W. Rajkumar, and S. Ramnath, “FABLE: Fairness-aware branching federated learning under events and concept drift,” Expert Systems with Applications, 2026, art. 131238, doi: 10.1016/j.eswa.2026.131238.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Allwine (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.
