Research map / Learning theory and algorithms / Matrix and tensor methods
Online Generalized Sparse Regression: How Does Overparametrization Help?
Shuoguang Yang, Qiang Sun
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 hard-thresholding steps K from 1 to 5 (ablation) | OHT · Online sparse regression, kappa=500, s=15 | last-iterate MSE: 0.529622 -> 0.070367 | Improved |
| increase hard-thresholding steps K from 1 to 20 (ablation) | OHT · Online sparse regression, kappa=500, s=14 | last-iterate MSE: 0.519131 -> 0.046217 | Improved |
| increase K from 5 to 20 at large projection sparsity (ablation) | OHT · Online sparse regression, kappa=500, s=40 | last-iterate MSE: 0.012848 -> 0.014919 | No clear effect |
| over-select projection sparsity from s=15 to s=20 (ablation) | OHT_K=5 · Online sparse regression, kappa=500 | last-iterate MSE: 0.070367 -> 0.011384 | Improved |
| under-pick projection rank s=3 instead of s=5 (ablation) | OHT_K=20 · Online low-rank matrix sensing | last-iterate MSE: 0.0499 -> 10.8299 | Hurt |
| under-pick projection rank s=4 instead of s=5 (ablation) | OHT_K=1 · Online low-rank matrix sensing | last-iterate MSE: 0.0546 -> 5.6856 | Hurt |
| increase projection rank from s=5 to s=20 (ablation) | OHT_K=5 · Online low-rank matrix sensing | last-iterate MSE: 0.0508 -> 0.2499 | Hurt |
| increase K from 1 to 20 in low-rank sensing (ablation) | OHT · Online low-rank matrix sensing, rank s=5 | last-iterate MSE: 0.0546 -> 0.0499 | No clear effect |
| use OHT instead of OS_LASSO at small initial batch (t0<=150) (ablation) | OHT · Online sparse regression, weak signal, independent and Toeplitz | last-iterate MSE | Improved |
| use OHT instead of OS_LASSO under strong signal (ablation) | OHT · Online sparse regression, strong signal | last-iterate MSE | No clear effect |
| use K=5,10,20 instead of K=1 at s<=30 (no initial batch) (ablation) | OHT · Online sparse regression, weak signal, no initial batch | last-iterate MSE | Improved |
Closest papers by meaning
- Distributed Sparse Regression via Penalization (NeurIPS 2024 · similarity 0.86)
- Communication-efficient Sparse Regression (JMLR 2017 · similarity 0.86)
- Efficient Sublinear-Regret Algorithms for Online Sparse Linear Regression with Limited Observation (NeurIPS 2017 · similarity 0.85)
- Adaptive Forward Stepwise: A Method for High Sparsity Regression (JMLR 2026 · similarity 0.85)
- Sharp Generalization for Nonparametric Regression by Over-Parameterized Neural Networks: A Distribution-Free Analysis in Spherical Covariate (ICML 2025 · similarity 0.83)
- Estimating Generalization Performance Along the Trajectory of Proximal SGD in Robust Regression (NeurIPS 2024 · 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.
