Research map / Learning theory and algorithms / Kernels and regression theory
A Consistent Information Criterion for Support Vector Machines in Diverging Model Spaces
Xiang Zhang, Yichao Wu, Lan Wang, Runze Li
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 |
|---|---|---|---|
| use SVMICH with high-dimensional penalty (Ln = log(log(n))) instead of SVMICL | SVM information criterion · Simulation Model 1 (n=200, p=2000) | Test error (%): 30.7 -> 25.4 | Improved |
| use SVMICH (Ln = sqrt(p)/log(n)) instead of SVMICL | SVM information criterion · Simulation Model 1 (n=400, p=4000) | Correct model selection (%): 4.0 -> 95.5 | Improved |
| use SVMICH (Ln = sqrt(p)/log(n)) instead of SVMICL | SVM information criterion · Simulation Model 2 (n=400, p=4000) | Correct model selection (%): 9.0 -> 89.0 | Improved |
| use more aggressive Ln = log(n) instead of sqrt(p)/log(n) (ablation) | SVMICH · Simulation Model 2 (n=400, p=4000) | Test error (%): 14.5 -> 20.1 | Hurt |
| use Ln = n^-1/3 instead of sqrt(p)/log(n) at small sample size (ablation) | SVMICH · Simulation Model 1 (n=200, p=2000) | Test error (%): 26.1 -> 29.7 | Hurt |
| increase sample size from n=200 to n=400 (p=10n) for Ln = log(n) (ablation) | SVMICH · Simulation Model 1 | Test error (%): 29.6 -> 22.4 | Improved |
| use SVMICH (Ln = sqrt(p)/log(n)) instead of 5-CV for tuning | SVMICH(lambda) · Tuning parameter selection (n=200, p=2000) | Test error (%): 7.6 -> 7.0 | Improved |
| use SVMICH (Ln = sqrt(p)/log(n)) instead of 5-CV Adj. | SVMICH(lambda) · Tuning parameter selection (n=200, p=2000) | Correct model selection (%): 67 -> 97 | Improved |
| use SVMICH (Ln = sqrt(p)/log(n)) instead of SVMICL for tuning | SVMICH(lambda) · MAQC-II breast cancer | Test error (%): 18.5 -> 9.9 | Improved |
| use SVMICH (Ln = log(n)) instead of 5-CV | SVMICH(lambda) · MAQC-II breast cancer | Test error (%): 10.8 -> 9.6 | Improved |
| use Chen&Chen criterion instead of SVMICH (Ln = sqrt(p)/log(n)) | SVMICH · Simulation Model 1 (n=300, p=3000) | Test error (%): 22.1 -> 22.1 | No clear effect |
Closest papers by meaning
- Asymptotic behavior of Support Vector Machine for spiked population model (JMLR 2017 · similarity 0.84)
- Support vector machines and linear regression coincide with very high-dimensional features (NeurIPS 2021 · similarity 0.84)
- Model-free Variable Selection in Reproducing Kernel Hilbert Space (JMLR 2016 · similarity 0.83)
- Statistical Rates of Convergence for Functional Partially Linear Support Vector Machines for Classification (JMLR 2022 · similarity 0.83)
- A Rigorous Information-Theoretic Definition of Redundancy and Relevancy in Feature Selection Based on (Partial) Information Decomposition (JMLR 2023 · similarity 0.83)
- Variational Information Maximization for Feature Selection (NeurIPS 2016 · 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.
