Research map / Label-efficient and robust learning / Recognition and long-tail
Similarity Search for Efficient Active Learning and Search of Rare Concepts
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 |
|---|---|---|---|
| Restrict candidate examples to nearest neighbors of labeled examples | MaxEnt active learning · ImageNet | mAP: 0.695 -> 0.695 | No clear effect |
| Restrict candidate examples to nearest neighbors of labeled examples | MaxEnt active search · ImageNet | recall: 57.2% -> 56.9% | No clear effect |
| Restrict candidate examples to nearest neighbors of labeled examples | MLP active learning · ImageNet | mAP: 0.693 -> 0.692 | No clear effect |
| Restrict candidate examples to nearest neighbors of labeled examples | MLP active search · ImageNet | recall: 74.5% -> 74.2% | No clear effect |
| Restrict candidate examples to nearest neighbors of labeled examples | ID active learning · ImageNet | mAP: 0.688 -> 0.688 | No clear effect |
| Restrict candidate examples to nearest neighbors of labeled examples | ID active search · ImageNet | recall: 50.8% -> 50.9% | No clear effect |
| Restrict candidate examples to nearest neighbors of labeled examples | MaxEnt active learning · OpenImages | mAP: 0.399 -> 0.386 | Hurt |
| Restrict candidate examples to nearest neighbors of labeled examples | MLP active learning · OpenImages | mAP: 0.398 -> 0.386 | Hurt |
| Increase the nearest-neighbor pool size to k=1,000 (ablation) | MaxEnt-SEALS and MLP-SEALS · OpenImages | mAP: improving mAP to 0.395 | Helped |
| Restrict candidate examples to nearest neighbors of labeled examples | Active learning and search · OpenImages | recall: 35.0% -> 35.1% | No clear effect |
| Restrict candidate examples to nearest neighbors of labeled examples | MaxEnt active learning · 10B images | mAP: 0.508 -> 0.504 | No clear effect |
| Use nearest-neighbor-restricted selection instead of scanning the entire pool | MaxEnt-SEALS · ImageNet | total time (seconds): 45.23 -> 12.49 | Helped |
Closest papers by meaning
- SIMILAR: Submodular Information Measures Based Active Learning In Realistic Scenarios (NeurIPS 2021 · similarity 0.87)
- Active Learning for Convolutional Neural Networks: A Core-Set Approach (ICLR 2018 · similarity 0.86)
- A Cross-Domain Benchmark for Active Learning (NeurIPS 2024 · similarity 0.86)
- Composite Active Learning: Towards Multi-Domain Active Learning with Theoretical Guarantees (AAAI 2024 · similarity 0.85)
- Plug and Play Active Learning for Object Detection (CVPR 2024 · similarity 0.85)
- Exploiting Distribution Constraints for Scalable and Efficient Image Retrieval (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.