Research map / Learning theory and algorithms / Sample complexity
Multiclass Learning from Contradictions
Sauptik Dhar, Vladimir Cherkassky, Mohak Shah
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 universum samples via MU-SVM instead of M-SVM | multiclass SVM · GTSRB | test error (%): 7.24 -> 5.53 | Improved |
| add universum samples via MU-SVM instead of M-SVM | multiclass SVM · ABCDETC | test error (%): 27.5 -> 22.1 | Improved |
| add universum samples via MU-SVM instead of M-SVM | multiclass SVM · ISOLET | test error (%): 3.6 -> 2.83 | Improved |
| direct multiclass universum formulation instead of OVA ECOC U-SVM | MU-SVM · ABCDETC | test error (%): 26.1 -> 22.1 | Improved |
| span-bound (Theorem 4) model selection instead of 5-fold CV | MU-SVM · GTSRB U3 | model selection time (x10^4 sec): 2.9 -> 0.9 | Improved |
| span-bound model selection instead of 5-fold CV | MU-SVM · GTSRB U3 | test error (%): 5.5 -> 5.2 | No clear effect |
| choice of universum: priority road sign (U3) vs others (U2) (ablation) | MU-SVM · GTSRB | test error (%): 7.4 -> 5.5 | Improved |
| choice of universum: random averaging (U4) vs uppercase letters (U1) (ablation) | MU-SVM · ABCDETC | test error (%): 26.1 -> 22.1 | Improved |
| increase number of universum samples beyond 500 (ablation) | MU-SVM · all datasets | test error | No clear effect |
Closest papers by meaning
- Multiclass learning with margin: exponential rates with no bias-variance trade-off (ICML 2022 · similarity 0.85)
- Surprises in Proper Positive-Only Learning (NeurIPS 2026 · similarity 0.84)
- A Trichotomy for List Transductive Online Learning (ICML 2025 · similarity 0.83)
- Improved Sample Complexity for Multiclass PAC Learning (NeurIPS 2024 · similarity 0.83)
- Balancing the Scales: A Theoretical and Algorithmic Framework for Learning from Imbalanced Data (ICML 2025 · similarity 0.83)
- Statistical Query Hardness of Multiclass Linear Classification with Random Classification Noise (ICML 2025 · 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.
