Research map / Learning theory and algorithms / Kernels and regression theory
Asymptotic behavior of Support Vector Machine for spiked population model
Hanwen Huang
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 |
|---|---|---|---|
| decrease soft-margin SVM tuning parameter tau toward a small threshold (ablation) | Soft-margin SVM under spiked model · Simulated spiked data (alpha=1, mu=2, sigma=2, lambda1=2, N=100) | cos(theta) and test error | Improved |
| increase tau to large values (ablation) | Soft-margin SVM · Simulated spiked data (alpha=1, mu=2, sigma=2, lambda1=2, N=100) | cos(theta) | Hurt |
| increase signal size mu (ablation) | Hard-margin SVM · Theory, alpha and lambda1 fixed | cos(theta) | Improved |
| increase sample ratio alpha (more training samples) (ablation) | Hard-margin SVM · Theory, mu and lambda1 fixed | cos(theta) | Improved |
| increase spike strength lambda1 (ablation) | Hard-margin SVM · Theory, alpha=1 | cos(theta) | Hurt |
| make class covariances unequal (heteroscedastic) (ablation) | Spiked-model SVM theory · Simulated, alpha=1, mu=2, sigma=2, lambda1=2 | agreement of theory with simulation | No clear effect |
| number of non-unit eigenvalues K increased from 10% to 30% of N (ablation) | Spiked-model SVM theory · Simulated, alpha=1, mu=2, sigma=2 | usefulness of cos(theta) for choosing tau: K=0.1N accurate -> K=0.3N bigger cos theta not always smaller error | Hurt |
| number of non-unit eigenvalues K increased to 50% of N (ablation) | Spiked-model SVM theory · Simulated, alpha=1, mu=2, sigma=2 | estimation of cos(theta) | Hurt |
Closest papers by meaning
- Support vector machines and linear regression coincide with very high-dimensional features (NeurIPS 2021 · similarity 0.87)
- Solving L1-regularized SVMs and Related Linear Programs: Revisiting the Effectiveness of Column and Constraint Generation (JMLR 2022 · similarity 0.85)
- A Consistent Information Criterion for Support Vector Machines in Diverging Model Spaces (JMLR 2016 · similarity 0.84)
- Safe Sample Screening for Robust Support Vector Machine (AAAI 2020 · similarity 0.84)
- Long-time asymptotics of noisy SVGD outside the population limit (ICLR 2025 · similarity 0.83)
- Faster Algorithms for Structured Linear and Kernel Support Vector Machines (ICLR 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.
