Research map / Label-efficient and robust learning / Self-supervised and contrastive
Does Learning from Decentralized Non-IID Unlabeled Data Benefit from Self Supervision?
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 |
|---|---|---|---|
| Pretrain with Dec-SSL instead of Dec-SLRep | ImageNet-100 representations transferred to MS-COCO · MS-COCO | bounding-box mAP: 19.8 -> 22.1 | Helped |
| Pretrain with Dec-SSL instead of Dec-SLRep | ImageNet-100 representations transferred to MS-COCO · MS-COCO | mask mAP: 19.7 -> 20.7 | Helped |
| Pretrain with Dec-SSL rather than train from scratch | MS-COCO detection · MS-COCO | bounding-box mAP: 20.5 -> 22.1 | Helped |
| Pretrain with Dec-SSL rather than train from scratch | MS-COCO segmentation · MS-COCO | mask mAP: 19.4 -> 20.7 | Helped |
| Use decentralized rather than centralized SSL pretraining | Amazon package segmentation · Amazon package segmentation (100% finetuning data) | mask mAP: 61.6 -> 61.2 | No clear effect |
| Use decentralized rather than centralized SSL pretraining | Amazon package segmentation · Amazon package segmentation (10% finetuning data) | mask mAP: 60.4 -> 60.1 | No clear effect |
| Use decentralized rather than centralized SSL pretraining | Amazon package segmentation · Amazon package segmentation (1% finetuning data) | mask mAP: 49.5 -> 48.8 | No clear effect |
| Use Dec-SSL rather than train from scratch | Amazon package segmentation · Amazon package segmentation (1% finetuning data) | mask mAP: 47.0 -> 48.8 | Helped |
| Use feature alignment and clustering in FeatARC | Dec-SSL · CIFAR-10 | classification accuracy | Helped |
| Use feature alignment in FeatARC (ablation) | Dec-SSL · CIFAR-10 | classification accuracy | Hurt |
| Use clustering in FeatARC (ablation) | Dec-SSL · CIFAR-10 | classification accuracy | Hurt |
| Increase local update epochs or reduce participant ratio | Dec-SSL compared with Dec-SLRep · ImageNet-100 under high non-IIDness | downstream classification accuracy | Mixed |
Closest papers by meaning
- DePRL: Achieving Linear Convergence Speedup in Personalized Decentralized Learning with Shared Representations (AAAI 2024 · similarity 0.87)
- SemiDFL: A Semi-Supervised Paradigm for Decentralized Federated Learning (AAAI 2025 · similarity 0.86)
- Combating Data Imbalances in Federated Semi-supervised Learning with Dual Regulators (AAAI 2024 · similarity 0.85)
- Continuous Contrastive Learning for Long-Tailed Semi-Supervised Recognition (NeurIPS 2024 · similarity 0.85)
- Self-supervised Representation Learning from Random Data Projectors (ICLR 2024 · similarity 0.85)
- InterLUDE: Interactions between Labeled and Unlabeled Data to Enhance Semi-Supervised Learning (ICML 2024 · similarity 0.85)
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.