Research map / Federated learning and privacy / Personalized and heterogeneous FL
Fine-Tuning Global Model via Data-Free Knowledge Distillation for Non-IID Federated Learning
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| fine-tune the aggregated global model with data-free distillation from local models | SCAFFOLD aggregation · CIFAR10, β=0.3 | test accuracy (%): 82.14±1.20 -> 84.38±0.49 | Helped |
| fine-tune the aggregated global model with data-free distillation from local models | SCAFFOLD aggregation · CIFAR100, β=0.3 | test accuracy (%): 54.36±0.32 -> 55.96±0.39 | Helped |
| add hard sample mining (ablation) | FedFTG · CIFAR10, β=0.3, 500 rounds | test accuracy (%): 82.49±0.42 -> 83.43±0.10 | Helped |
| add customized label sampling (ablation) | FedFTG · CIFAR10, β=0.3, 500 rounds | test accuracy (%): 82.39±0.22 -> 83.43±0.10 | Helped |
| add class-level ensemble (ablation) | FedFTG · CIFAR10, β=0.3, 500 rounds | test accuracy (%): 82.40±0.21 -> 83.43±0.10 | Helped |
| add hard sample mining, customized label sampling, and class-level ensemble (ablation) | FedFTG · CIFAR10, β=0.3, 500 rounds | test accuracy (%): 81.98±0.14 -> 83.43±0.10 | Helped |
| add classification loss (ablation) | FedFTG · CIFAR10, β=0.3, 500 rounds | test accuracy (%): 82.50±0.55 -> 83.43±0.10 | Helped |
| add diversity loss (ablation) | FedFTG · CIFAR10, β=0.3, 500 rounds | test accuracy (%): 82.52±0.35 -> 83.43±0.10 | Helped |
| add classification and diversity losses (ablation) | FedFTG · CIFAR10, β=0.3, 500 rounds | test accuracy (%): 82.12±0.16 -> 83.43±0.10 | Helped |
| replace KL divergence with mean square error (ablation) | FedFTG · CIFAR10, β=0.3, 500 rounds | test accuracy (%): 83.43±0.10 -> 10.17±0.26 | Hurt |
| generate pseudo data at feature level instead of input level (ablation) | FedFTG · CIFAR10, β=0.3 | test accuracy (%): 84.38±0.49 -> 82.76±0.81 | Hurt |
| add server-side FedFTG fine-tuning | FedAvg · CIFAR10, β=0.3 | test accuracy (%): 79.59±1.01 -> 82.27±0.67 | Helped |
Closest papers by meaning
- Efficient Federated Learning via Clients-to-Server Knowledge Distillation (Student Abstract) (AAAI 2025 · similarity 0.91)
- Feature Distillation is the Better Choice for Model-Heterogeneous Federated Learning (NeurIPS 2025 · similarity 0.91)
- Overcoming Data and Model heterogeneities in Decentralized Federated Learning via Synthetic Anchors (ICML 2024 · similarity 0.91)
- Fake It Till Make It: Federated Learning with Consensus-Oriented Generation (ICLR 2024 · similarity 0.90)
- An Aggregation-Free Federated Learning for Tackling Data Heterogeneity (CVPR 2024 · similarity 0.90)
- Data-Free Black-Box Federated Learning via Zeroth-Order Gradient Estimation (AAAI 2025 · similarity 0.90)
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.