Research map / Federated learning and privacy / Personalized and heterogeneous FL
Federated Semi-Supervised Learning with Inter-Client Consistency & Disjoint Learning
Read the paperCheck 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 |
|---|---|---|---|
| Use FedMatch with inter-client consistency and parameter decomposition instead of a naive federated SSL combination | federated semi-supervised learning · CIFAR-10 Batch-IID and Batch-NonIID, labels-at-client and labels-at-server | accuracy (%): 52.13 vs 47.45; 44.95 vs 25.61; 52.25 vs 46.31; 44.17 vs 26.01 | Helped |
| Use FedMatch instead of the strongest listed federated SSL baseline | Fashion-MNIST Streaming-NonIID, labels-at-client · Fashion-MNIST Streaming-NonIID, labels-at-client | accuracy (%): 73.71 -> 77.95 | Helped |
| Use FedMatch instead of the strongest listed federated SSL baseline | Fashion-MNIST Streaming-NonIID, labels-at-server · Fashion-MNIST Streaming-NonIID, labels-at-server | accuracy (%): 83.34 -> 84.15 | Helped |
| Add inter-client consistency loss (ablation) | FedMatch · CIFAR-10 Batch-IID, 100 clients | accuracy | Helped |
| Add parameter decomposition for supervised and unsupervised learning (ablation) | FedMatch · CIFAR-10 Batch-IID | accuracy (%): 47.51 -> 54.5 | Helped |
| Remove the supervised parameter component sigma (ablation) | FedMatch parameter decomposition · CIFAR-10 Batch-IID | accuracy | Hurt |
| Remove the unsupervised parameter component psi (ablation) | FedMatch parameter decomposition · CIFAR-10 Batch-IID | accuracy | Hurt |
| Increase labeled examples per class from 5 to 10 with FedMatch (ablation) | FedMatch · CIFAR-10 Batch-IID | accuracy (%): 54.5 -> 60.65 | Helped |
| Increase labeled examples per class from 5 to 10 without decomposition (ablation) | FedMatch without decomposition · CIFAR-10 Batch-IID | accuracy (%): 47.51 -> 51.15 | Helped |
| Increase labeled examples per class from 5 to 10 (ablation) | FedProx-UDA · CIFAR-10 Batch-IID | accuracy (%): 47.45 -> 41.4 | Hurt |
| Increase labeled examples per class from 5 to 10 (ablation) | FedProx-FixMatch · CIFAR-10 Batch-IID | accuracy (%): 47.2 -> 34.25 | Hurt |
Closest papers by meaning
- (FL)$^2$: Overcoming Few Labels in Federated Semi-Supervised Learning (NeurIPS 2024 · similarity 0.95)
- Progressive Distribution Matching for Federated Semi-Supervised Learning (AAAI 2025 · similarity 0.94)
- Mind the Gap: Confidence Discrepancy Can Guide Federated Semi-Supervised Learning Across Pseudo-Mismatch (CVPR 2025 · similarity 0.91)
- SemiDFL: A Semi-Supervised Paradigm for Decentralized Federated Learning (AAAI 2025 · similarity 0.91)
- Combating Data Imbalances in Federated Semi-supervised Learning with Dual Regulators (AAAI 2024 · similarity 0.91)
- BSemiFL: Semi-supervised Federated Learning via a Bayesian Approach (ICML 2025 · similarity 0.91)
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.