Research map / Learning theory and algorithms / Prediction and decision losses
Model-free Change-Point Detection Using AUC of a Classifier
Rohit Kanrar, Feiyu Jiang, Zhanrui Cai
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 the AUC statistic instead of the CUSUM statistic on vgg16 output (ablation) | vgg16 · CIFAR10 change-point, Cat -> Dog | ARI: 0.002 -> 0.965 | Improved |
| use the AUC statistic instead of the CUSUM statistic on vgg19 output (ablation) | vgg19 · CIFAR10 change-point, Cat -> Dog | ARI: 0 -> 0.963 | Improved |
| use the AUC statistic instead of the CUSUM statistic on vgg16 output (ablation) | vgg16 · CIFAR10 no change-point, Deer | rejection rate at 5% level: 0% -> 4% | Improved |
| use classifier-based AUC (vgg16) instead of graph-based gseg_wei | vgg16 · CIFAR10 change-point, Cat -> Dog | ARI: 0.378 -> 0.965 | Improved |
| use classifier-based AUC (vgg16) instead of changeforest | vgg16 · CIFAR10 change-point, Cat -> Dog | ARI: 0.952 -> 0.965 | Improved |
| use regularized logistic classifier inside AUC statistic (ablation) | Logis · high-dimensional distribution and covariance changes | ARI | Hurt |
| use fully connected neural network classifier inside AUC statistic (ablation) | Fnn · high-dimensional mean and distribution changes | ARI | Mixed |
| use random forest classifier inside AUC statistic (ablation) | Rf · high-dimensional changes of all types | ARI | Improved |
| use the Logis classifier instead of Hddc distance-based test for computation | Logis · Dense Mean Change, p=500 | computation time (s): 0.101 vs >10^4 | Improved |
| use random forest AUC instead of changeforest | Rf · Dense Mean Change, p=500 | computation time (s): 4.258 -> 0.697 | Improved |
| calibrate test with simulated Brownian quantiles | changeAUC · Standard, banded, exponential nulls | type-I error at 5% level: 2.5 to 5.7 | Improved |
Closest papers by meaning
- Random Forests for Change Point Detection (JMLR 2023 · similarity 0.85)
- Optimal Online Change Detection via Random Fourier Features (NeurIPS 2025 · similarity 0.85)
- Reducing sequential change detection to sequential estimation (ICML 2024 · similarity 0.83)
- AUC Maximization under Positive Distribution Shift (NeurIPS 2024 · similarity 0.83)
- Positive-unlabeled AUC Maximization under Covariate Shift (ICML 2025 · similarity 0.82)
- The Sample Complexity of Multiple Change Point Identification under Bandit Feedback (NeurIPS 2026 · similarity 0.82)
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.
