Research map / Label-efficient and robust learning / Semi-supervised learning
Optimizing Data Collection for Machine Learning
Rafid Mahmood, James Lucas, Jose M. Alvarez, Sanja Fidler, Marc T. Law
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 LOC instead of Power Law Regression | Data collection policy · CIFAR-10, T=5 | failure rate: 86% -> 29% | Improved |
| Use LOC instead of Power Law Regression | Data collection policy · CIFAR-100, T=5 | failure rate: 48% -> 2% | Improved |
| Use LOC instead of Power Law Regression | Data collection policy · ImageNet, T=5 | failure rate: 56% -> 2% | Improved |
| Use LOC instead of Power Law Regression | Data collection policy · BDD100K segmentation, T=5 | failure rate: 23% -> 0% | Improved |
| Use LOC instead of Power Law Regression | Data collection policy · nuScenes, T=5 | failure rate: 62% -> 0% | Improved |
| Use LOC instead of Power Law Regression | Data collection policy · VOC, T=5 | failure rate: 6% -> 0% | Improved |
| Use LOC instead of Power Law Regression | Two-source collection policy · CIFAR-100 two data types, T=5, cost (0.01, 0.001) | failure rate: 35% -> 24% | Improved |
| Use LOC instead of Power Law Regression | Two-source collection policy · BDD100K semi-supervised, T=5, cost (1, 0.01) | failure rate: 21% -> 2% | Improved |
| Use LOC instead of Power Law Regression | Two-source collection policy · CIFAR-100 two data types, T=5, cost (0.01, 0.001) | cost ratio: 1.22 -> 0.79 | Improved |
| Increase collection rounds from T=1 to T=5 | LOC · CIFAR-100 | failure rate: 4% -> 2% | Improved |
| Increase collection rounds from T=1 to T=5 | LOC · BDD100K segmentation | failure rate: 12% -> 0% | Improved |
| Increase collection rounds from T=1 to T=5 | LOC · VOC | failure rate: 25% -> 0% | Improved |
Closest papers by meaning
- Swift Sampler: Efficient Learning of Sampler by 10 Parameters (NeurIPS 2024 · similarity 0.85)
- Compute-Constrained Data Selection (ICLR 2025 · similarity 0.84)
- Repeated Random Sampling for Minimizing the Time-to-Accuracy of Learning (ICLR 2024 · similarity 0.84)
- Lightweight Dataset Pruning without Full Training via Example Difficulty and Prediction Uncertainty (ICML 2025 · similarity 0.84)
- GIO: Gradient Information Optimization for Training Dataset Selection (ICLR 2024 · similarity 0.84)
- Enhancing Semi-Supervised Learning via Representative and Diverse Sample Selection (NeurIPS 2024 · similarity 0.84)
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.
