Research map / Label-efficient and robust learning / Recognition and long-tail
Generalized Resubstitution for Classification Error Estimation
Parisa Ghane, Ulisses Braga-Neto
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 3NN posterior-probability estimator instead of zero bootstrap | error estimation, linear SVM · Synthetic data, n=20 | computation time (ms): 138.75 -> 2.31 | Improved |
| use spherical bolstered resubstitution instead of zero bootstrap | error estimation, RBF SVM · Synthetic data, n=100 | computation time (ms): 434.10 -> 61.53 | Improved |
| use 10-fold cross-validation instead of spherical bolstered resubstitution | error estimation, 3-NN · Synthetic data, n=100 | computation time (ms): 343.50 -> 13.11 | Improved |
| add bolstering kernels to 3NN posterior-probability estimator | 3NNpp, 3-NN classifier · Synthetic data, n=20 | computation time (ms): 4.00 -> 136.46 | Hurt |
| increase sample size from n=20 to n=100 for spherical bolstered resubstitution | spherical bolstered resubstitution, linear SVM · Synthetic data | computation time (ms): 9.65 -> 41.06 | Hurt |
| replace plain resubstitution with bolstered resubstitution | resubstitution error estimator · Synthetic data, linear and CART classifiers | bias and RMS of error estimate | Improved |
| use bolstered resubstitution on 3NN classifier | bolstered resubstitution · Synthetic data, 3NN classifier | bias of error estimate | No clear effect |
| use semi-bolstered resubstitution instead of bolstered resubstitution | 3NN classifier error estimation · Synthetic data, 3NN classifier | RMS of error estimate | Improved |
| use bolstered posterior-probability estimator instead of cross-validation and bootstrap | error estimation · Synthetic data, non-3NN classifiers | RMS of error estimate | Improved |
| use Naive-Bayes bolstered resubstitution instead of plain resubstitution | LeNet-5 error estimation · MNIST subsets n=200-800 | bias of error estimate | Improved |
| use 11-NN posterior-probability and bolstered posterior-probability estimators instead of plain resubstitution | LeNet-5 error estimation · MNIST subsets n=200-800 | bias of error estimate | Hurt |
| add data-driven kappa calibration to Naive-Bayes bolstered resubstitution | Naive-Bayes bolstered resubstitution · MNIST subsets n=200-800 | bias and RMS of error estimate | Improved |
Closest papers by meaning
- Generalized Resubstitution for Regression Error Estimation (JMLR 2026 · similarity 0.88)
- Learning to Reject Meets Long-tail Learning (ICLR 2024 · similarity 0.84)
- ResAD: A Simple Framework for Class Generalizable Anomaly Detection (NeurIPS 2024 · similarity 0.84)
- A Data-Driven Measure of Relative Uncertainty for Misclassification Detection (ICLR 2024 · similarity 0.83)
- DiffAug: A Diffuse-and-Denoise Augmentation for Training Robust Classifiers (NeurIPS 2024 · similarity 0.83)
- REDUCR: Robust Data Downsampling using Class Priority Reweighting (NeurIPS 2024 · similarity 0.83)
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.
