Research map / Federated learning and privacy / Communication-efficient FL
Learning To Collaborate in Decentralized Learning of Personalized Models
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 |
|---|---|---|---|
| Use validation-loss-optimized neighbor mixing weights for personalized decentralized learning | L2C · CIFAR-10 | test accuracy: 88.72±0.29 -> 90.14±0.34 | Helped |
| Use validation-loss-optimized neighbor mixing weights for personalized decentralized learning | L2C · CIFAR-100 | test accuracy: 52.44±0.70 -> 59.00±0.42 | Helped |
| Use validation-loss-optimized neighbor mixing weights for personalized decentralized learning | L2C · MiniImageNet | test accuracy: 44.56±0.77 -> 50.03±0.75 | Helped |
| Generate mixing weights with an attention mechanism over model-update representations | meta-L2C · CIFAR-10 | test accuracy: 88.72±0.29 -> 92.10±0.72 | Helped |
| Generate mixing weights with an attention mechanism over model-update representations | meta-L2C · CIFAR-100 | test accuracy: 52.44±0.70 -> 58.28±1.26 | Helped |
| Generate mixing weights with an attention mechanism over model-update representations | meta-L2C · MiniImageNet | test accuracy: 44.56±0.77 -> 48.80±1.42 | Helped |
| Pretrain meta-L2C on CIFAR-100 tasks and transfer it to CIFAR-10 tasks without fine-tuning | meta-L2C · CIFAR-10 | test accuracy: 92.10±0.72 -> 91.33±0.89 | No clear effect |
| Pretrain meta-L2C on CIFAR-10 tasks and transfer it to CIFAR-100 tasks without fine-tuning | meta-L2C · CIFAR-100 | test accuracy: 58.28±1.26 -> 56.59±1.40 | No clear effect |
| Vary the training-to-validation data ratio (ablation) | L2C and meta-L2C · CIFAR-10 | average test accuracy | No clear effect |
| Vary the per-node communication budget (ablation) | L2C and meta-L2C · CIFAR-10 | average test accuracy | No clear effect |
| Vary the per-node communication budget (ablation) | DPSGD · CIFAR-10 | average test accuracy | Helped |
Closest papers by meaning
- DePRL: Achieving Linear Convergence Speedup in Personalized Decentralized Learning with Shared Representations (AAAI 2024 · similarity 0.86)
- DIMAT: Decentralized Iterative Merging-And-Training for Deep Learning Models (CVPR 2024 · similarity 0.84)
- Decentralized Directed Collaboration for Personalized Federated Learning (CVPR 2024 · similarity 0.83)
- SP-CACW: Convergence-Aware Client Weighting for Selfish Personalized Learning (NeurIPS 2026 · similarity 0.83)
- Learning to Specialize: Joint Gating-Expert Training for Adaptive MoEs in Decentralized Settings (NeurIPS 2025 · similarity 0.83)
- ZipIt! Merging Models from Different Tasks without Training (ICLR 2024 · similarity 0.83)
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.