@article{Zhang2026, 
author = {Yu-Pei Zhang and Rou-Jia Feng and Yi-Fei Wang and Shu-Hui Liu and Xue-Qun Shang},
title = {Federated Deep Subspace Clustering},
year = {2026},
journal = {Journal of Computer Science and Technology},
volume = {41},
number = {2},
pages = {609-620},
keywords = {deep learning, federated learning, image clustering, deep subspace clustering, private protection},
url = {https://www.sciopen.com/article/10.1007/s11390-025-5304-4},
doi = {10.1007/s11390-025-5304-4},
abstract = {This paper presents Federated Deep Subspace Clustering (FDSC), a privacy-preserving deep subspace clustering model built upon a federated learning framework. In FDSC, each client employs a dedicated deep subspace clustering network to process its locally isolated data. This network consists of an encoder, a self-expressive layer, and a decoder. To enable collaboration across clients, the encoder network is shared with a central server, allowing communication and model aggregation. Furthermore, FDSC enhances local clustering performance by preserving the neighborhood relationships among data samples within each client. By integrating federated learning with locality preservation, the encoder learns more expressive features, which in turn improve the self-expressiveness and clustering accuracy. Extensive experiments on public datasets show that FDSC outperforms existing methods, benefiting from both federated learning and locality preservation.}
}