Research map / Optimization / Implicit bias and generalization
Implicit Regularization in Over-Parameterized Support Vector Machine
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 initialization size alpha from 1e-4 to 1e-10 (ablation) | over-parameterized GD SVM · Simulated sparse SVM (n=200, p=400) | relative estimation error | Helped |
| decrease initialization size alpha (ablation) | over-parameterized GD SVM · Simulated sparse SVM | convergence speed | Hurt |
| implicit regularization via GD instead of explicit L1 (Lasso SVM) | over-parameterized GD SVM · Varying signal strength m | estimation error | Helped |
| GD instead of Lasso SVM | over-parameterized GD SVM · Varying signal strength m | false positive error | Helped |
| GD instead of Lasso SVM | over-parameterized GD SVM · Varying signal strength m | test prediction accuracy | No clear effect |
| GD instead of Lasso SVM | over-parameterized GD SVM · Varying sample size n=50 to 400 | estimation, prediction, variable selection | Helped |
| GD instead of Lasso SVM | over-parameterized GD SVM · Complex signal structures A-E | prediction and variable selection | Helped |
| sample x from t(3) heavy-tailed distribution instead of Gaussian (ablation) | over-parameterized GD SVM · Complex signal structures A-E | performance | No clear effect |
| increase smoothness parameter gamma over 2.5e-5 to 1e-3 (ablation) | over-parameterized GD SVM · Default simulation | estimation error | No clear effect |
| GD vs Lasso on uniform-input data, structures A-E | over-parameterized GD SVM · Uniform [-1,1] inputs | estimation accuracy and variable selection error | Helped |
| GD vs oracle on prediction under uniform inputs | over-parameterized GD SVM · Uniform [-1,1] inputs | prediction accuracy | No clear effect |
| alternative data generating schemes Model 1 and Model 2 (ablation) | over-parameterized GD SVM · Model 1 and Model 2 | estimation error and prediction accuracy | No clear effect |
Closest papers by meaning
- Implicit Gradient Regularization (ICLR 2021 · similarity 0.86)
- Scaling Up Sparse Support Vector Machines by Simultaneous Feature and Sample Reduction (ICML 2017 · similarity 0.85)
- Deep Principal Support Vector Machines for Nonlinear Sufficient Dimension Reduction (ICML 2025 · similarity 0.84)
- Improving Adaptivity via Over-Parameterization in Sequence Models (NeurIPS 2024 · similarity 0.84)
- Generalization in Kernel Regression Under Realistic Assumptions (ICML 2024 · similarity 0.83)
- Implicit Regularization for Tubal Tensor Factorizations via Gradient Descent (ICML 2025 · similarity 0.83)
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.