Research map / Federated learning and privacy / Personalized and heterogeneous FL
FedFA: Federated Feature Augmentation
Tianfei Zhou, Ender Konukoglu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Adding FedFA (federation-aware feature statistic augmentation) to FedAvg | FedAvg on Office-Caltech 10 · Office-Caltech 10 | Top-1 accuracy average (%): +4.6 over FedAvg (78.5 to 83.1) | Improved |
| Adding FedFA to FedAvg | FedAvg on DomainNet · DomainNet | Top-1 accuracy average (%): +3.7 over FedAvg (62.8 to 66.5) | Improved |
| Adding FedFA to FedAvg | FedAvg on CIFAR-10 and EMNIST · CIFAR-10 Dir(0.6), Dir(0.3), EMNIST | Top-1 accuracy (%): +3.0 (Dir 0.6), +2.7 (Dir 0.3), +2.9 (EMNIST) over FedAvg | Improved |
| Adding FedFA to FedAvg in federated domain generalization to unseen clients | FedAvg · Office-Caltech 10, DomainNet, ProstateMRI leave-one-client-out | average accuracy/Dice (%): +3.9 Office, +2.2 DomainNet (approx from table averages 60.4->64.3, 47.5->51.2), +5.2 ProstateMRI (72.3->77.5) | Improved |
| Random statistic perturbation with fixed std (FedFA-R) instead of federation-aware variances (ablation) | FedAvg · DomainNet and ProstateMRI | accuracy / Dice (%): DomainNet 62.8 to 61.0; ProstateMRI 86.5 to 86.1; Office no change (78.5 to 78.6) | Hurt |
| Using client-specific variances only (FedFA-C) instead of full federation-aware variances (ablation) | FedFA · Office-Caltech 10, DomainNet, ProstateMRI | accuracy / Dice (%): Office 83.1 to 79.5; DomainNet 66.5 to 63.7; ProstateMRI 89.5 to 87.8 | Hurt |
| Direct fusion of client-specific and client-sharing variances instead of adaptive (t-distribution) fusion (ablation) | FedFA · Office-Caltech 10, DomainNet, ProstateMRI | accuracy / Dice (%): Office 83.1 to 80.6; DomainNet 66.5 to 64.1; ProstateMRI 89.5 to 86.9 | Hurt |
| Adding FedFA on top of conventional augmentation (MixStyle, Mixup, M-Mixup, MoEx) (ablation) | FedAvg with traditional augmentation · Office-Caltech 10, DomainNet, ProstateMRI | accuracy / Dice (%): e.g. Office MixStyle 79.9 to FedFA+MixStyle 84.0; ProstateMRI 88.5 to 90.2 | Improved |
| Applying FFA to more convolutional stages (set {1,2,3,4,5} vs single stage) (ablation) | FedFA · Office-Caltech 10, DomainNet, ProstateMRI | accuracy / Dice (%): Single stage {1}: Office 78.8, DomainNet 63.5, ProstateMRI 88.5; all five stages: 83.1, 66.5, 89.5 | Improved |
| Adding FedFA to FedAvg in mini-ProstateMRI small-data setting | FedAvg · mini-ProstateMRI (1/6 of training data) | Dice (%): +3.0 over FedAvg | Improved |
| Adding FedFA to FedAvg with local data reduced to 1/6 (mini-ProstateMRI) | FedAvg · mini-ProstateMRI | Dice (%): +3.0 (absolute values not given in text) | Improved |
Later papers that compared with this method
3 later accepted papers report FedFA as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- FedSDA: Federated Stain Distribution Alignment for Non-IID Histopathological Image Classification (AAAI 2026 · better on 3 of 3 shared results)
- Enhancing Visual Representation with Textual Semantics: Textual Semantics-Powered Prototypes for Heterogeneous Federated Learning (CVPR 2026 · better on 0 of 1 shared result)
- A Simple Data Augmentation for Feature Distribution Skewed Federated Learning (CVPR 2025 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FedMix — FedMix: Approximation of Mixup under Mean Augmented Federated Learning (ICLR 2021) · better on 7 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- A Simple Data Augmentation for Feature Distribution Skewed Federated Learning (CVPR 2025 · similarity 0.88)
- Personalized Federated Learning via Feature Distribution Adaptation (NeurIPS 2024 · similarity 0.87)
- FedLoGe: Joint Local and Generic Federated Learning under Long-tailed Data (ICLR 2024 · similarity 0.87)
- Tackling Feature-Classifier Mismatch in Federated Learning via Prompt-Driven Feature Transformation (NeurIPS 2025 · similarity 0.86)
- Flexible Sharpness-Aware Personalized Federated Learning (AAAI 2025 · similarity 0.86)
- FedQS: Optimizing Gradient and Model Aggregation for Semi-Asynchronous Federated Learning (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
