Research map / Learning theory and algorithms / Kernels and regression theory
On Regularizing Rademacher Observation Losses
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 |
|---|---|---|---|
| Add SLOPE regularizer (ablation) | Omega-R.AdaBoost · 20 domains | domains won vs unregularized: 1 -> 19 | Helped |
| Add regularization (any) (ablation) | Omega-R.AdaBoost · Magic | test error: more than 15% | Helped |
| Add regularization (any) (ablation) | Omega-R.AdaBoost · Kaggle Give me some credit | test error: more than 20% | Helped |
| Use Omega-R.AdaBoost with regularizer instead of AdaBoost | Omega-R.AdaBoost · 20 domains | test error: 10-11 of 20 domains won vs best Ada | Helped |
| Choose best vs worst regularizer | Omega-R.AdaBoost · Statlog | test error: exceeds 15% | Mixed |
| Subsample rados heavily (n=10000) | Omega-R.AdaBoost · Marketing, Kaggle | test error | No clear effect |
| Add regularization (ablation) | Omega-R.AdaBoost · Spectf, Transfusion, Hill-noise, Winered, Magic, Marketing | sparsity | Helped |
| Use ridge vs unregularized (ablation) | Omega-R.AdaBoost · 20 domains | domains won: 3 -> 17 | Helped |
| Use lasso vs unregularized (ablation) | Omega-R.AdaBoost · 20 domains | domains won: 3 -> 17 | Helped |
| Use l-infinity vs unregularized (ablation) | Omega-R.AdaBoost · 20 domains | domains won: 2 -> 18 | Helped |
| Use SLOPE vs lasso (ablation) | Omega-R.AdaBoost · 20 domains | domains won: 10 vs 4 | Helped |
| Use rados instead of examples on large domains | Omega-R.AdaBoost vs AdaBoost · Hardware, Kaggle | convergence | Helped |
Closest papers by meaning
- iFlood: A Stable and Effective Regularizer (ICLR 2022 · similarity 0.85)
- Optimizing Nondecomposable Data Dependent Regularizers via Lagrangian Reparameterization Offers Significant Performance and Efficiency Gains (AAAI 2020 · similarity 0.85)
- Learning Optimal Linear Regularizers (ICML 2019 · similarity 0.85)
- Beyond Tikhonov: faster learning with self-concordant losses, via iterative regularization (NeurIPS 2021 · similarity 0.85)
- Regularized Gradient Boosting (NeurIPS 2019 · similarity 0.85)
- Multiclass Loss Geometry Matters for Generalization of Gradient Descent in Separable Classification (NeurIPS 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.