Research map / Label-efficient and robust learning / Self-supervised and contrastive
Self-supervised Learning from a Multi-view Perspective
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 |
|---|---|---|---|
| Add inverse predictive loss to contrastive learning (ablation) | Contrastive SSL · Omniglot test set | 1-nearest-neighbor accuracy | Helped |
| Add inverse predictive loss to contrastive learning (ablation) | SimCLR · CIFAR10 | test performance | Helped |
| Use contrastive learning instead of forward predictive learning (ablation) | SSL objectives · Omniglot test set | test accuracy | Helped |
| Use contrastive learning instead of forward predictive learning (ablation) | SSL objectives · Omniglot training | epochs to best performance | Helped |
| Train contrastive learning for long epochs (ablation) | Contrastive learning · Omniglot test set | test accuracy | Hurt |
| Combine contrastive and forward predictive losses (ablation) | SSL objectives · Omniglot test set | test performance | Helped |
| Use a pre-trained ResNet with BERT rather than a pre-trained ResNet alone (ablation) | Visual-textual SSL · MS COCO validation | downstream classification performance | Helped |
| Pre-train the ResNet rather than relying on BERT pre-training (ablation) | Visual-textual SSL · MS COCO validation | downstream classification performance | Helped |
| Add inverse predictive loss to contrastive learning (ablation) | Visual-textual SSL · MS COCO validation | downstream classification performance | Helped |
| Use reverse binary cross-entropy reconstruction loss rather than MSE or BCE (ablation) | Forward predictive learning · Visual representation learning | test generalization performance | Helped |
| Construct self-supervised signals from another example of the same character, differing in drawing style and augmentations (ablation) | Contrastive SSL · Omniglot test set | test generalization performance | Helped |
Closest papers by meaning
- Learning Representations by Maximizing Mutual Information Across Views (NeurIPS 2019 · similarity 0.90)
- Revisiting Self-Supervised Visual Representation Learning (CVPR 2019 · similarity 0.89)
- When Does Contrastive Visual Representation Learning Work? (CVPR 2022 · similarity 0.89)
- Self-Supervised Learning as Discrete Communication (ICML 2026 · similarity 0.88)
- Improving Self-Supervised Learning by Characterizing Idealized Representations (NeurIPS 2022 · similarity 0.88)
- No More Shortcuts: Realizing the Potential of Temporal Self-Supervision (AAAI 2024 · similarity 0.86)
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.