Research map / Federated learning and privacy / Personalized and heterogeneous FL
Model-Contrastive 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 |
|---|---|---|---|
| add model-contrastive loss to local training | FedAvg · CIFAR-10, CIFAR-100, and Tiny-Imagenet non-IID test sets | top-1 accuracy: 66.3% -> 69.1%; 64.5% -> 67.5%; 23.0% -> 25.1% | Helped |
| add model-contrastive loss to local training | FedAvg · CIFAR-10, CIFAR-100, and Tiny-Imagenet | communication rounds to reach FedAvg's standard-round accuracy: 100 -> 27; 100 -> 43; 20 -> 11 | Helped |
| replace model-contrastive loss with ℓ2 representation regularization (ablation) | MOON · CIFAR-10, CIFAR-100, and Tiny-Imagenet | top-1 accuracy: 69.1% -> 65.8%; 67.5% -> 66.9%; 25.1% -> 24.0% | Hurt |
| add projection head (ablation) | MOON · CIFAR-10 and CIFAR-100 | top-1 accuracy: 66.8% -> 69.1%; 66.1% -> 67.5% | Helped |
| reduce contrastive-loss weight to μ=0.1 (ablation) | MOON · CIFAR-10 | top-1 accuracy: 66.3% -> 66.5% | No clear effect |
| increase contrastive-loss weight from μ=0 to μ=1 (ablation) | MOON · CIFAR-10 | top-1 accuracy: 66.3% -> 68.4% | Helped |
| increase maximum negative pairs from 1 to 2 (ablation) | MOON · CIFAR-10 | top-1 accuracy: 69.1% -> 67.2% | Hurt |
| increase maximum negative pairs from 1 to 5 (ablation) | MOON · CIFAR-10 | top-1 accuracy: 69.1% -> 67.7% | Hurt |
| increase maximum negative pairs from 1 to 100 (ablation) | MOON · CIFAR-10 | top-1 accuracy: 69.1% -> 63.5% | Hurt |
| apply model-contrastive loss with IID rather than heterogeneous party data (ablation) | FedAvg · CIFAR-10 IID partition | top-1 accuracy: 73.4% -> 73.6% (μ=0.1); 73.4% -> 73.6% (μ=1) | No clear effect |
Closest papers by meaning
- Relaxed Contrastive Learning for Federated Learning (CVPR 2024 · similarity 0.90)
- A Mutual Information Perspective on Federated Contrastive Learning (ICLR 2024 · similarity 0.86)
- FedQS: Optimizing Gradient and Model Aggregation for Semi-Asynchronous Federated Learning (NeurIPS 2025 · similarity 0.85)
- FedTGP: Trainable Global Prototypes with Adaptive-Margin-Enhanced Contrastive Learning for Data and Model Heterogeneity in Federated Learning (AAAI 2024 · similarity 0.84)
- FedCS: Coreset Selection for Federated Learning (CVPR 2025 · similarity 0.84)
- Tackling Feature-Classifier Mismatch in Federated Learning via Prompt-Driven Feature Transformation (NeurIPS 2025 · similarity 0.84)
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.