Research map / Federated learning and privacy / Personalized and heterogeneous FL
Personalized Semantics Excitation for Federated Image Classification
Haifeng Xia, Kai Li, Zhengming Ding
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 adaptive personalized channel excitation and cross-model attention exchange to local customization | personalized FL (Ditto/FedRep-style) · CIFAR-100 (50 clients, 15 classes) | average accuracy: 58.94 -> 62.46 (vs FedRep) | Improved |
| remove adaptive channel excitation module (Ours-v1) (ablation) | PSE · CIFAR-100 (50,15); FEMNIST (200,3), (200,4) | test accuracy | Improved |
| add cross-model attention exchange (vs removing it, Ours-v2) (ablation) | PSE · CIFAR-100 (50,15); FEMNIST (200,3), (200,4) | test accuracy | Improved |
| add learnable position information in cross-model attention (ablation) | PSE · CIFAR-100 (50,15); FEMNIST (200,3), (200,4) | test accuracy | Improved |
| replace Grad-CAM with Score-CAM for channel importance (ablation) | PSE · FEMNIST and FashionMNIST (Gray, Color) | average accuracy: 88.81 -> 88.63 (FEMNIST 200,3); 89.58 -> 89.77 (FashionMNIST 100,3) | Mixed |
| replace Grad-CAM with Grad-CAM++ for channel importance (ablation) | PSE · FEMNIST and FashionMNIST (Gray, Color) | average accuracy: 87.86 -> 87.35 (FEMNIST 200,4); 88.12 -> 88.86 (FashionMNIST 100,4) | No clear effect |
| use PSE instead of FedAvg+FT or FedRep for client customization | personalized FL · FashionMNIST (Gray, 100 clients, 4 classes) | std of client accuracy: 9.02% / 8.49% -> 5.01% | Improved |
| naive local fine-tuning (Local) vs methods using global model | client models · FEMNIST (Gray, Color), 200 clients, 3 classes | average accuracy: Local 81.97 vs PSE 88.81 | Hurt |
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.
- Ditto — Ditto: Fair and Robust Federated Learning Through Personalization (ICML 2021) · better on 4 of 4 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
- Personalized Federated Learning via Feature Distribution Adaptation (NeurIPS 2024 · similarity 0.90)
- Flexible Sharpness-Aware Personalized Federated Learning (AAAI 2025 · similarity 0.89)
- Tackling Feature-Classifier Mismatch in Federated Learning via Prompt-Driven Feature Transformation (NeurIPS 2025 · similarity 0.89)
- pFedES: Generalized Proxy Feature Extractor Sharing for Model Heterogeneous Personalized Federated Learning (AAAI 2025 · similarity 0.88)
- FedLoGe: Joint Local and Generic Federated Learning under Long-tailed Data (ICLR 2024 · similarity 0.87)
- Fine-Tuning Personalization in Federated Learning to Mitigate Adversarial Clients (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.
