Research map / Federated learning and privacy / Personalized and heterogeneous FL
Required Spine Optional Limbs: Heterogeneous Federated Learning via Backbone-sharing and Activation-guided Selection
Mingsheng Cao, Hongliang Chen, Ming Hu, Fei Gao, Qiaolong Ding, Wenke Huang, Xiaofei Xie, Junlong Zhou
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 activation-guided neuron selection (vs random selection) (ablation) | SpineFL · CIFAR-10 (IID) | accuracy (%): 85.25 -> 86.67 | Improved |
| add the shared backbone mechanism (ablation) | SpineFL · CIFAR-10 (IID) | accuracy (%): 83.15 -> 86.67 | Improved |
| SpineFL vs FedDSE, most skewed data | SpineFL, ResNet-34 · CIFAR-10 (Non-IID α=0.3) | accuracy (%): 76.64 -> 78.30 | Improved |
| SpineFL vs FedDSE, most skewed data | SpineFL, VGG-16 · CIFAR-10 (Non-IID α=0.3) | accuracy (%): 78.69 -> 84.34 | Improved |
| SpineFL vs FedDSE, EMNIST | SpineFL, ResNet-34 · EMNIST (Non-IID α=0.3) | accuracy (%): 96.06 -> 98.14 | Improved |
| SpineFL vs FedDSE, low-resource configuration | SpineFL · CIFAR-10 (low-resource) | accuracy (%): 76.78 -> 77.83 | Improved |
| SpineFL vs FedDSE, high-resource configuration | SpineFL · CIFAR-10 (high-resource) | accuracy (%): 82.69 -> 84.32 | Improved |
| use more selected devices per round (K=50 vs K=10) | SpineFL · CIFAR-10 (α=0.3) | accuracy: K=50 achieves 0.8 accuracy within 400 rounds, k=10 converges to about 0.83 after 1000 rounds | Mixed |
| increase total devices from 100 to 500 | SpineFL · CIFAR-10 (α=0.3) | convergence rounds | Mixed |
Closest papers by meaning
- Data-Free Black-Box Federated Learning via Zeroth-Order Gradient Estimation (AAAI 2025 · similarity 0.84)
- FIARSE: Model-Heterogeneous Federated Learning via Importance-Aware Submodel Extraction (NeurIPS 2024 · similarity 0.84)
- SpaFL: Communication-Efficient Federated Learning With Sparse Models And Low Computational Overhead (NeurIPS 2024 · similarity 0.82)
- FedDAT: An Approach for Foundation Model Finetuning in Multi-Modal Heterogeneous Federated Learning (AAAI 2024 · similarity 0.82)
- DapperFL: Domain Adaptive Federated Learning with Model Fusion Pruning for Edge Devices (NeurIPS 2024 · similarity 0.82)
- Ranking-based Client Imitation Selection for Efficient Federated Learning (ICML 2024 · similarity 0.82)
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.
