Research map / Federated learning and privacy / Communication-efficient FL
Aggregating Capacity in FL through Successive Layer Training for Computationally-Constrained Devices
Kilian Pfeiffer, Ramin Khalili, Joerg Henkel
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 |
|---|---|---|---|
| use SLT instead of small model baseline | FL training · ResNet20/CIFAR10 IID, s=0.125 | accuracy (%): 66.3 -> 74.1 | Improved |
| use SLT instead of FedRolex | FL training · ResNet20/CIFAR10 IID, s=0.125 | accuracy (%): 21.7 -> 74.1 | Improved |
| use SLT instead of small model on easy dataset | FL training · ResNet20/FEMNIST IID, s=0.125 | accuracy (%): 82.3 -> 84.4 | No clear effect |
| use small model instead of full model | FL training · CIFAR100 and TinyImageNet IID | accuracy (%): up to 25.4 pp | Hurt |
| use SLT instead of small model at s=0.66 | FL training · DenseNet40/CIFAR100 IID, s=0.66 | accuracy (%): 55.9 -> 53.3 | Hurt |
| use SLT instead of small model under non-IID | FL training · ResNet20/CIFAR10 non-IID, s=0.125 | accuracy (%): 44.7 -> 52.4 | Improved |
| use SLT instead of small model | FL training · ResNet44/TinyImageNet IID, s=0.125 | accuracy (%): 19.8 -> 33.5 | Improved |
| use SLT instead of best baseline under heterogeneous constraints | FL training · ResNet44/TinyImageNet heterogeneous [0.125, 0.25, 0.5, 1.0] | accuracy (%): 27.5 -> 35.9 | Improved |
| add a fourth constraint level for HeteroFL | HeteroFL · ResNet44/TinyImageNet heterogeneous | accuracy (%): 24.1 -> 23.3 | Hurt |
| add a third constraint level (1.0) for FjORD | FjORD · DenseNet40/CIFAR100 heterogeneous | accuracy (%): 38.7 -> 36.9 | Hurt |
| use SLT instead of small model on ImageNet 64x64 | FL training · ResNet56/ImageNet IID, s=0.125 | accuracy (%): 8.9 -> 24.2 | Improved |
| use fewer random subsets in FD (limit randomness) (ablation) | FD · ResNet20/CIFAR10, s=0.25 | accuracy (%) | 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.
- FedRolex — FedRolex: Model-Heterogeneous Federated Learning with Rolling Sub-Model Extraction (NeurIPS 2022) · better on 8 of 8 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
- WHALE-FL: Wireless and Heterogeneity Aware Latency Efficient Federated Learning over Mobile Devices via Adaptive Subnetwork Scheduling (AAAI 2025 · similarity 0.89)
- Low Precision Local Training is Enough for Federated Learning (NeurIPS 2024 · similarity 0.89)
- SparsyFed: Sparse Adaptive Federated Learning (ICLR 2025 · similarity 0.88)
- FedASMU: Efficient Asynchronous Federated Learning with Dynamic Staleness-Aware Model Update (AAAI 2024 · similarity 0.88)
- FedEL: Federated Elastic Learning for Heterogeneous Devices (NeurIPS 2025 · similarity 0.88)
- FedQS: Optimizing Gradient and Model Aggregation for Semi-Asynchronous Federated Learning (NeurIPS 2025 · 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.
