Research map / Federated learning and privacy / Personalized and heterogeneous FL
FedPAT: Federated Test-Time Adaptation via Prototype Affinity Topology
Shunxin Guo, Jiaqi Lv, Zhiqiang Kou, Shuxia Lin, Xin Geng
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 |
|---|---|---|---|
| add prototype affinity topology (PAT) modeling (ablation) | FedPAT · CIFAR-10-C severity 5 | avg accuracy: 64.22 -> 70.56 | Improved |
| add topology-aware non-parametric inference (TNI) (ablation) | FedPAT · CIFAR-10-C severity 5 | avg accuracy: 68.78 -> 70.56 | Improved |
| add parametric classifier prediction in the fusion (ablation) | FedPAT · CIFAR-10-C severity 5 | avg accuracy: 68.56 -> 70.56 | Improved |
| add topology-guided validation refinement (ablation) | FedPAT · CIFAR-10-C severity 5 | avg accuracy: 69.73 -> 70.56 | Improved |
| add PAT under label shift (ablation) | FedPAT · CIFAR-10-C severity 5, LS | accuracy: 72.66 -> 80.73 | Improved |
| use FedPAT instead of ATP | ATP · CIFAR-10-C severity 5 | avg accuracy: 69.54 -> 70.56 | Improved |
| use FedPAT instead of ATP | ATP · CIFAR-100-C severity 5, feature shift | accuracy: 39.81 -> 38.99 | Hurt |
| use FedPAT instead of ATP | ATP · Tiny-ImageNet-C severity 5 | avg accuracy: 43.96 -> 44.74 | Improved |
| use FedPAT instead of ATP with random severity | ATP · Tiny-ImageNet-C random severity | avg accuracy: 42.98 -> 45.24 | Improved |
| use FedPAT instead of ATP at lower severity | ATP · CIFAR-100-C severity 3 | avg accuracy: 45.92 -> 46.50 | Improved |
| use FedPAT instead of BN-Adapt | BN-Adapt · CIFAR-10-C severity 5 | avg accuracy: 64.22 -> 70.56 | Improved |
| use PAT similarity instead of feature distribution similarity as cross-client anchor | Cross-client consistency · CIFAR-10-C feature shift | mean similarity: 0.336 -> 0.875 | Improved |
Closest papers by meaning
- FLUX: Efficient Descriptor-Driven Clustered Federated Learning under Arbitrary Distribution Shifts (NeurIPS 2025 · similarity 0.89)
- Beyond the Federation: Topology-aware Federated Learning for Generalization to Unseen Clients (ICML 2024 · similarity 0.88)
- TTA-FedDG: Leveraging Test-Time Adaptation to Address Federated Domain Generalization (AAAI 2025 · similarity 0.88)
- Dual-Personalizing Adapter for Federated Foundation Models (NeurIPS 2024 · similarity 0.87)
- Efficient Adaptive Federated Optimization (NeurIPS 2025 · similarity 0.87)
- Efficient Federated Learning against Heterogeneous and Non-stationary Client Unavailability (NeurIPS 2024 · similarity 0.87)
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.
