to improve model performance by differently erasing each clients domain skew and enhancing their consensus. First， [Submitted on 17 Mar 2025] Title: Federated Learning with Domain Shift Eraser Authors: Zheng Wang， we propose a novel FL framework， and generalizability. Comments: Accepted by CVPR2025 Subjects: Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2503.13063 [cs.CV] (or arXiv:2503.13063v1 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2503.13063 Focus to learn more arXiv-issued DOI via DataCite ， a regularization term is applied to promise the effectiveness of feature deskewing by pulling local statistics of DSEs outputs close to the globally consistent ones. Finally， Federated Domain Shift Eraser (FDSE)， DFE modules are fairly aggregated and broadcast to all the clients to maximize their consensus， preventing the model from learning consistent representation space. In this paper， Cheng Wang View a PDF of the paper titled Federated Learning with Domain Shift Eraser， clients data originating from diverse domains may degrade model performance due to domain shifts，。
Zheng Wang， we formulate the model forward passing as an iterative deskewing process that extracts and then deskews features alternatively. This is efficiently achieved by decomposing each original layer in the neural network into a Domain-agnostic Feature Extractor (DFE) and a Domain-specific Skew Eraser (DSE). Then， and DSE modules are personalized for each client via similarity-aware aggregation to erase their domain skew differently. Comprehensive experiments were conducted on three datasets to confirm the advantages of our method in terms of accuracy， efficiency， Zihui Wang， Xiaoliang Fan， by Zheng Wang and 4 other authors View PDFHTML (experimental) Abstract: Federated learning (FL) is emerging as a promising technique for collaborative learning without local data leaving their devices. However。