Research map / Learning theory and algorithms / Kernels and regression theory
Accelerating Cross-Validation in Multinomial Logistic Regression with $\ell_1$-Regularization
Tomoyuki Obuchi, Yoshiyuki Kabashima
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 SAACV instead of 10-fold CV on a large problem | l1 multinomial logistic regression CV · ISOLET (N=617, M=6238, L=26) | computation time (seconds): 7825 -> 689 seconds | Improved |
| use ACV instead of 10-fold CV on a large problem | l1 multinomial logistic regression CV · ISOLET (N=617, M=6238, L=26) | computation time (seconds): 7825 -> 5173 seconds | Improved |
| increase feature dimensionality N from 50-100 to 400-800 (ablation) | ACV and SAACV approximation accuracy · Simulated data (L=8, alpha=2, rho0=0.5) | agreement with literal LOO CV: N=50,100 discriminable difference; N=400, 800 much smaller | Improved |
| add strong correlation in features (common components fraction 0.9 or noise correlation 0.9) (ablation) | SAACV · Simulated data (N=200, L=8) | approximation accuracy | No clear effect |
| add class heterogeneity by amplifying feature norms (Omega=100 or 10) (ablation) | SAACV · Simulated data and MNIST digits 0 and 1 | approximation accuracy vs LOO CV | Hurt |
| tighten convergence threshold delta from 1e-6 to 1e-8 (ablation) | literal CV vs approximations · Simulated data (N=400, L=8) | CV error curve: delta 10^-6 vs 10^-8 | Mixed |
| grow system size N beyond 400 (800 and above) (ablation) | SAACV vs ACV computation time · Simulated data (L=8, alpha=2, rho0=0.5) | computation time: for N>=400 approximate CV is better; for N>=800 ACV cost exceeds SAACV | Improved |
Closest papers by meaning
- Safe Grid Search with Optimal Complexity (ICML 2019 · similarity 0.85)
- Bayes beats Cross Validation: Efficient and Accurate Ridge Regression via Expectation Maximization (NeurIPS 2023 · similarity 0.85)
- Cross-regularization: Adaptive Model Complexity through Validation Gradients (ICML 2025 · similarity 0.85)
- Crossing the Validation crisis: Cross-validation reduces benchmarking variance surprisingly well (NeurIPS 2026 · similarity 0.84)
- A Provably Accurate Randomized Sampling Algorithm for Logistic Regression (AAAI 2024 · similarity 0.83)
- Benefits of Early Stopping in Gradient Descent for Overparameterized Logistic Regression (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.
