Research map / Label-efficient and robust learning / Recognition and long-tail
Unsupervised Feature Learning by Cross-Level Instance-Group Discrimination
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 cross-level discrimination (CLD) to MoCo | MoCo · VOC07+12 | AP: 55.9 -> 56.7 | Helped |
| add CLD to NPID (ablation) | NPID · CIFAR-10 | kNN top-1 accuracy: 80.8% -> 82.7% | Helped |
| use separate feature and group subspaces instead of a shared subspace (ablation) | NPID+CLD · CIFAR-10 | kNN top-1 accuracy: 82.7% -> 84.2% | Helped |
| compare across augmentations (ablation) | NPID+CLD with separate subspaces · CIFAR-10 | kNN top-1 accuracy: 84.2% -> 86.5% | Helped |
| set λ to 0.25 (ablation) | NPID+CLD · ImageNet-100 | linear top-1 accuracy: 75.3 -> 79.7 | Helped |
| increase λ from 0.25 to 3.0 (ablation) | NPID+CLD · ImageNet-100 | linear top-1 accuracy: 79.7 -> 76.6 | Hurt |
| set temperature T to 0.2 (ablation) | CLD · ImageNet-100 | linear top-1 accuracy: 79.3% -> 81.7% | Helped |
| use k-Means rather than spectral clustering with 10 groups (ablation) | CLD · Kitchen-HC | top-1 kNN accuracy: 77.1% -> 78.9% | Helped |
Closest papers by meaning
- Unsupervised Feature Learning via Non-Parametric Instance Discrimination (CVPR 2018 · similarity 0.88)
- On the Alignment Between Supervised and Self-Supervised Contrastive Learning (ICLR 2026 · similarity 0.87)
- Temperature Schedules for self-supervised contrastive methods on long-tail data (ICLR 2023 · similarity 0.87)
- Unsupervised Representation Learning by Invariance Propagation (NeurIPS 2020 · similarity 0.86)
- On the Similarities of Embeddings in Contrastive Learning (ICML 2025 · similarity 0.85)
- $\mathbb{X}$-Sample Contrastive Loss: Improving Contrastive Learning with Sample Similarity Graphs (ICLR 2025 · 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.