Research map / Learning theory and algorithms / Prediction and decision losses
Optimal Sample Selection Through Uncertainty Estimation and Its Application in Deep Learning
Yong Lin, Chen Liu, Chenlu Ye, Qing Lian, Yuan Yao, Tong Zhang
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 |
|---|---|---|---|
| add clipping of maximum uncertainty in the sampling ratio (COPS-clip) (ablation) | COPS-vanilla · CIFAR10, CIFARBinary, CIFAR10-N coreset selection and active learning | test accuracy | Improved |
| use vanilla uncertainty sampling instead of uniform sampling | uniform sampling · CIFARBinary-600-WithY | test accuracy (%): 88.31 -> 75.26 | Hurt |
| add clipping of maximum uncertainty (COPS-clip) under label corruption in the low-density region, threshold 3 or 10 times the minimum uncertainty (ablation) | COPS-vanilla on linear simulation · 2-D logistic simulation, corruption zeta(x1) = -3 | parameter error and regret loss | Improved |
| increase corruption of the low-density region (zeta from 0 to -3) (ablation) | COPS-vanilla · 2-D logistic simulation | parameter error and regret loss | Hurt |
| estimate uncertainty with independently trained models of different initializations instead of bootstrap (ablation) | COPS · CIFAR10-8,000 | test accuracy | Improved |
| estimate uncertainty with Monte Carlo dropout instead of different initializations (ablation) | COPS · CIFAR10 coreset selection (WithY) | test accuracy | No clear effect |
| use COPS instead of uniform sampling across ResNet56, MobileNetV2 and DenseNet121 | uniform sampling · CIFAR10 multiple architectures | test accuracy | Improved |
| use COPS instead of uniform sampling on SVHN, Places365 subset and IMDB | uniform sampling · SVHN, Places365 (subset), IMDB | test accuracy | Improved |
Closest papers by meaning
- Stable Coresets via Posterior Sampling: Aligning Induced and Full Loss Landscapes (NeurIPS 2025 · similarity 0.89)
- Refined Coreset Selection: Towards Minimal Coreset Size under Model Performance Constraints (ICML 2024 · similarity 0.87)
- Deep Active Learning with Noise Stability (AAAI 2024 · similarity 0.87)
- Repeated Random Sampling for Minimizing the Time-to-Accuracy of Learning (ICLR 2024 · similarity 0.86)
- A Bayesian Approach to Data Point Selection (NeurIPS 2024 · similarity 0.85)
- The Impact of Coreset Selection on Spurious Correlations and Group Robustness (NeurIPS 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.
