Research map / Learning theory and algorithms / Kernels and regression theory
Model Selection via the VC Dimension
Merlin Mpoudeu, Bertrand Clarke
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 |
|---|---|---|---|
| increase sample size from 600 to 700 with similar design points (ablation) | VCD estimator · Simulated linear model p=60 | estimated VC dimension (true 60): 50 -> 57 | Improved |
| choose design points covering the full range [0,n] (ablation) | VCD estimator · Simulated linear model p=60, n=2000 | estimated VC dimension (true 60): 50 -> 59 | Improved |
| use more design points over a small range instead of a wide range (ablation) | VCD estimator · Simulated linear model p=60, n=2000 | estimated VC dimension (true 60): 59 -> 50 | Hurt |
| raise n/p towards 27 from about 10 (ablation) | VCD estimator · Simulated linear models p=15 to 60 | discrimination between models: 10 -> 27 | Improved |
| better design points at n/p of 10 (ablation) | VCD estimator · Simulated linear model n/p = 10 | inference quality | Improved |
| use VCD instead of AIC, BIC, CV, ERM1, ERM2 | model selection · Simulated linear models, over-sized candidates | sensitivity to overfitting | Improved |
| use supremum within each interval instead of the mean in Algorithm 1 (ablation) | VCD estimator · Simulated linear models | estimator quality | Hurt |
| estimate VC dimension via ERM1/ERM2 instead of VCD directly | model selection · Simulated linear models, growing p at fixed n | robustness | Hurt |
Closest papers by meaning
- Model-free Variable Selection in Reproducing Kernel Hilbert Space (JMLR 2016 · similarity 0.85)
- Identifying a Minimal Class of Models for High--dimensional Data (JMLR 2017 · similarity 0.85)
- Joint Continuous and Discrete Model Selection via Submodularity (JMLR 2022 · similarity 0.84)
- Towards a statistical theory of data selection under weak supervision (ICLR 2024 · similarity 0.83)
- Distributionally Robust Feature Selection (NeurIPS 2025 · similarity 0.83)
- Differentially Private High-dimensional Variable Selection via Integer Programming (NeurIPS 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.
