Research map / Federated learning and privacy / Personalized and heterogeneous FL
Local or Global: Selective Knowledge Assimilation for Federated Learning with Limited Labels
Yae Jee Cho, Gauri Joshi, Dimitrios Dimitriadis
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 global-local consistency regularization (lambda_0 > 0 versus 0) (ablation) | FedLabel · CIFAR10, CIFAR100, OrganAMNIST, BloodMNIST | test accuracy: approximately 5-9% improvement | Improved |
| use very high confidence threshold beta=0.9 instead of lower values (ablation) | FedLabel · CIFAR10, CIFAR100, OrganAMNIST, BloodMNIST | test accuracy: lowering beta below 0.9 improves by at most approximately 17% | Hurt |
| use more local steps tau to obtain the local model (tau=5 is worst) (ablation) | FedLabel · CIFAR10 and others | test accuracy | Improved |
| use variance instead of entropy as confidence measure (ablation) | FedLabel · OrganAMNIST, BloodMNIST | test accuracy | Improved |
| use variance instead of entropy as confidence measure (ablation) | FedLabel · CIFAR10, CIFAR100 | test accuracy | No clear effect |
| add strong augmentation (RandAugment) to pseudo-labeled samples (ablation) | FedLabel · CIFAR10 | test accuracy | Improved |
| lower data heterogeneity (alpha 1 versus 0.1) | FedLabel vs baselines · CIFAR10, OrganAMNIST | accuracy gap to baselines: gap smaller by around 11-14% | Hurt |
| increase labeled fraction (50% vs 20% for CIFAR10) | FedLabel vs baselines · CIFAR10 | accuracy gap to baselines: gap 15-24% -> 8-15% | Hurt |
| apply FedLabel with 20% labels on CIFAR100 versus fully labeled FedProx | FedLabel · CIFAR100 | test accuracy: 24.87 -> 18.42 | Hurt |
| naively combine FixMatch with FedAvg under limited labels | FedAvg · CIFAR10 | test accuracy: 28.83 -> 23.99 | Hurt |
Closest papers by meaning
- (FL)$^2$: Overcoming Few Labels in Federated Semi-Supervised Learning (NeurIPS 2024 · similarity 0.92)
- BSemiFL: Semi-supervised Federated Learning via a Bayesian Approach (ICML 2025 · similarity 0.91)
- Mind the Gap: Confidence Discrepancy Can Guide Federated Semi-Supervised Learning Across Pseudo-Mismatch (CVPR 2025 · similarity 0.90)
- Combating Data Imbalances in Federated Semi-supervised Learning with Dual Regulators (AAAI 2024 · similarity 0.90)
- Exploring One-Shot Semi-supervised Federated Learning with Pre-trained Diffusion Models (AAAI 2024 · similarity 0.89)
- FedImpro: Measuring and Improving Client Update in Federated Learning (ICLR 2024 · similarity 0.89)
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.
