CVPR Findings2026
FedNPC: Stochastic Noise-driven Post-hoc Classifier Calibration Method for Federated Long-tailed Learning
Jintong Gao, He Zhao, Yibo Yang, Dandan Guo
Abstract
Federated long-tailed learning (Fed-LT) simultaneously tackles data heterogeneity across clients and global longtailed distribution challenges. Existing approaches often depend on auxiliary datasets, gradients, or statistical information, which not only increase communication overhead but also introduce significant privacy leakage risks. To address these issues, we first conduct a systematic empirical study of the classifier bias under imbalanced data. We observe that the norm imbalance problem is substantially more severe in Fed-LT than in the centralized LT setting. Motivated by this observation, we propose FedNPC (stochastic Noise-driven Post-hoc classifier Calibration), a novel data-free post-processing calibration method for Fed-LT, using a virtual, balanced dataset. Furthermore, we provide theoretical insights showing that isotropic noise can effectively mitigate classifier norm imbalance. Extensive experiments on multiple Fed-LT benchmarks demonstrate the superior performance and generalization ability of FedNPC. Our code is available at https://github.com/JintongGao/FedNPC.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- Long-tail learning via logit adjustment cites Decoupling Representation and Classifier for Long-Tailed Recognition
- Federated Learning on Non-IID Data Silos: An Experimental Study cites Ensemble Distillation for Robust Model Fusion in Federated Learning
- Federated Learning on Non-IID Data Silos: An Experimental Study cites Tackling the Objective Inconsistency Problem in Heterogeneous Federated Optimization
Builds on21
- Tackling the Objective Inconsistency Problem in Heterogeneous Federated OptimizationJianyu Wang, Qinghua Liu, Hao Liang, Gauri Joshi et al.NeurIPS 2020 · 2,231 citations
- Ensemble Distillation for Robust Model Fusion in Federated LearningTao Lin, Lingjing Kong, Sebastian U. Stich, Martin JaggiNeurIPS 2020 · 1,615 citations
- Decoupling Representation and Classifier for Long-Tailed RecognitionBingyi Kang, Saining Xie, Marcus Rohrbach, Zhicheng Yan et al.ICLR 2020 · 1,496 citations
- Federated Learning on Non-IID Data Silos: An Experimental StudyQinbin Li, Yiqun Diao, Quan Chen, Bingsheng HeICDE 2022 · 1,110 citations
- Long-tail learning via logit adjustmentAditya Krishna Menon, Sadeep Jayasumana, Ankit Singh Rawat, Himanshu Jain et al.ICLR 2021 · 937 citations
Related papers
- Fed-GraB: Federated Long-tailed Learning with Self-Adjusting Gradient BalancerZikai Xiao, Zihan Chen, Songshang Liu, Hualiang Wang et al.NeurIPS 2023 · 39 citations
- FedCART: Tackling Long-Tailed Distributions in Federated Adversarial Training via Classifier RefinementYuchen Qin, Yizhi Zhou, Junxiao Wang, Xin Xie et al.CVPR 2026
- Global Balanced Experts for Federated Long-Tailed LearningYaopei Zeng, Lei Liu, Li Liu, Li Shen et al.ICCV 2023 · 10 citations
- Fine-Tuning Impairs the Balancedness of Foundation Models in Long-tailed Personalized Federated LearningShihao Hou, Chikai Shang, Zhiheng Yang, Jiacheng Yang et al.CVPR 2026 · 2 citations
- FedTail-DT: A Dual-Teacher Framework for Long-Tailed Heterogeneous FL with CLIP Prototypes and Adaptive AggregationZijie Guo, Jinghua Zhu, Gang Du, Kejia ZhangKDD 2026
