Research map / Learning theory and algorithms / Prediction and decision losses
Leaky Hockey Stick Loss: The First Negatively Divergent Margin-based Loss Function for Classification
Oh-Ran Kwon, Hui Zou
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 |
|---|---|---|---|
| replace hinge loss with LHS loss | Kernel SVM · Simulation Ex2, n=200, Gaussian kernel | misclassification rate (%): 16.38 -> 15.45 | Improved |
| replace hinge loss with LHS loss | Kernel SVM · Simulation Ex2, n=900, linear kernel | misclassification rate (%): 38.11 -> 37.12 | Improved |
| replace hinge loss with LHS loss | Kernel SVM · Simulation Ex1, n=50, linear kernel | misclassification rate (%): 34.35 -> 34.63 | No clear effect |
| replace hinge loss with LHS loss | Kernel SVM · Simulation Ex3, n=50, Gaussian kernel | misclassification rate (%): 30.97 -> 31.36 | Hurt |
| replace hinge loss with LHS loss | Kernel SVM · UCI Sonar, linear kernel | misclassification rate (%): 25.10 -> 23.39 | Improved |
| replace hinge loss with LHS loss | Kernel SVM · UCI Banknote, Gaussian kernel | misclassification rate (%): 0.01 -> 0.44 | Hurt |
| replace hinge loss with LHS loss | Kernel SVM · UCI Musk, Gaussian kernel | misclassification rate (%): 8.29 -> 10.48 | Hurt |
| replace hinge loss with LHS loss | Kernel SVM · UCI datasets (15), linear kernel | misclassification rate (%): outperforms on ten data sets | Improved |
| replace hinge loss with LHS loss | Kernel SVM · UCI datasets (15), Gaussian kernel | misclassification rate (%): outperforms on six data sets | Mixed |
| replace hinge loss with LHS loss | Kernel SVM · UCI Australian, linear kernel computation time (s) | misclassification rate (%): 344.94 -> 0.35 | Improved |
| replace hinge loss with LHS loss | Kernel SVM · UCI Chess, Gaussian kernel | misclassification rate (%): 0.93 -> 1.70 | Hurt |
| increase sample size from 50 to 900 | LHS classifier · Simulation Ex1, Gaussian LHS classifier | misclassification rate (%): 17.11 -> 11.47 | Improved |
Closest papers by meaning
- Unified Binary and Multiclass Margin-Based Classification (JMLR 2024 · similarity 0.87)
- Support Vector Machine Classifier via Soft-Margin Loss (TPAMI 2021 · similarity 0.86)
- The Nyström method for convex loss functions (JMLR 2024 · similarity 0.86)
- Weston-Watkins Hinge Loss and Ordered Partitions (NeurIPS 2020 · similarity 0.85)
- Label Distributionally Robust Losses for Multi-class Classification: Consistency, Robustness and Adaptivity (ICML 2023 · similarity 0.85)
- EnsLoss: Stochastic Calibrated Loss Ensembles for Preventing Overfitting in Classification (ICML 2025 · 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.
