Research map / Learning theory and algorithms / Kernels and regression theory
Revisiting minimum description length complexity in overparameterized models
Raaz Dwivedi, Chandan Singh, Bin Yu, Martin Wainwright
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 |
|---|---|---|---|
| tune ridge lambda with Prac-MDL-COMP instead of LOOCV on 1199_BNG_echoMonths | ridge regression · PMLB datasets, d/n=1 | test MSE: 1.47 -> 1.00 | Improved |
| tune ridge lambda with Prac-MDL-COMP instead of LOOCV on 218_house_8L | ridge regression · PMLB datasets, d/n=1 | test MSE: 4.37 -> 1.01 | Improved |
| tune ridge lambda with Prac-MDL-COMP instead of LOOCV on 344_mv | ridge regression · PMLB datasets, d/n=1 | test MSE: 0.85 -> 1.06 | Hurt |
| tune ridge lambda with Prac-MDL-COMP instead of LOOCV on 201_pol | ridge regression · PMLB datasets, d/n=1 | test MSE: 0.82 -> 0.87 | Hurt |
| use Prac-MDL-COMP instead of LOOCV when data is very limited | ridge regression · PMLB datasets | test MSE | Improved |
| increase training samples (lower d/n) (ablation) | Prac-MDL-COMP vs Ridge-CV · PMLB datasets | test MSE | No clear effect |
| select lambda with Prac-MDL-COMP instead of BIC or ARD | ridge regression · fMRI voxels | test MSE | Improved |
| select lambda with Prac-MDL-COMP instead of LOOCV | ridge regression · fMRI voxels | test MSE | Mixed |
| replace linear ridge with NTK kernel ridge under Prac-MDL-COMP selection | Prac-MDL-COMP · fMRI voxels | test MSE | No clear effect |
| use Prac-MDL-COMP tuned ridge instead of OLS when d is near n | linear regression · Gaussian simulation, n=100 | test MSE | Improved |
Closest papers by meaning
- Prediction Under Imperfect Compression: A Theory of Approximate MDL (NeurIPS 2026 · similarity 0.86)
- Data Mixture Optimization: A Multi-fidelity Multi-scale Bayesian Framework (NeurIPS 2025 · similarity 0.86)
- Measuring Stochastic Data Complexity with Boltzmann Influence Functions (ICML 2024 · similarity 0.85)
- No Free Lunch from Random Feature Ensembles: Scaling Laws and Near-Optimality Conditions (ICML 2025 · similarity 0.85)
- Does Compression Imply Generalization? A Minimum Description Length Perspective (NeurIPS 2026 · similarity 0.85)
- Power Lines: Scaling laws for weight decay and batch size in LLM pre-training (NeurIPS 2025 · similarity 0.85)
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.
