Geometric Manifold Rectification for Imbalanced Learning
Xubin Wang, Qing Li, Weijia Jia, [0.4em] BNU-BNBU Institute of Artificial Intelligence, Future NetworksBeijing Normal-Hong Kong Baptist University and Beijing Normal University at Zhuhai · 2026
Read Geometric Manifold Rectification for Imbalanced Learning by Xubin Wang, Qing Li, Weijia Jia, [0.4em] BNU-BNBU Institute of Artificial Intelligence, Future NetworksBeijing Normal-Hong Kong Baptist University and Beijing Normal University at Zhuhai (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-13045.txt.
Description
Imbalanced data trip up classifiers when the majority class intrudes into the minority's territory, blurring the edge, most of all with noise and overlap. Xubin Wang and two co-authors propose shape surface rectification, a system that exploits local shape instead of symmetric cleaning rules. It guesses confidence from inverse-distance-weighted nearest-neighbour voting with an adaptive metric, capturing local reliability.
Cleaning is then asymmetric: strict on majority samples, protective of minority samples through a cap on minority removal. That asymmetry avoids the common failure of deleting informative minority points along with noise. Tests on multiple benchmarks show the method competing with strong sampling plain rivals. For tabular imbalanced learning, shape-aware cleaning offers a more careful other to blanket undersampling.
Catalog record
| CREATOR | Xubin Wang, Qing Li, Weijia Jia, [0.4em] BNU-BNBU Institute of Artificial Intelligence, Future NetworksBeijing Normal-Hong Kong Baptist University and Beijing Normal University at Zhuhai |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 74 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-13045