Research map / Optimization / Implicit bias and generalization
Asymptotics of Stochastic Gradient Descent with Dropout Regularization in Linear Models
Jiaqi Li, Johannes Schmidt-Hieber, Wei Biao Wu
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 |
|---|---|---|---|
| raise learning rate slightly above 2/lambda_max(X) | GD with dropout · n,d=100,5, p=0.9 | empirical contraction constant: 0.97 -> 1.02 | Hurt |
| raise learning rate slightly above 2/lambda_max(X) | GD with dropout · n,d=100,100, p=0.9 | empirical contraction constant: 0.93 -> 1.15 | Hurt |
| increase number of SGD steps n | online ASGD dropout CI · d=3, p=0.9, alpha=0.01 | coverage probability: 0.815 -> 0.945 | Improved |
| increase number of SGD steps n | online ASGD dropout CI · d=50, p=0.5, alpha=0.0125 | coverage probability: 0.735 -> 0.925 | Improved |
| lower retaining probability p from 0.9 to 0.5 | online ASGD dropout CI · d=50, alpha=0.01 | coverage probability: 0.950 -> 0.910 | Hurt |
| increase learning rate from 0.01 to 0.1 | online ASGD dropout CI · d=3, p=0.5 | coverage probability: 0.925 -> 0.950 | Improved |
Closest papers by meaning
- Dropout Regularization Versus l2-Penalization in the Linear Model (JMLR 2024 · similarity 0.91)
- Stochastic Modified Equations and Dynamics of Dropout Algorithm (ICLR 2024 · similarity 0.89)
- Nonasymptotic Analysis of Stochastic Gradient Descent with the Richardson–Romberg Extrapolation (ICLR 2025 · similarity 0.88)
- Statistical Guarantees for High-Dimensional Stochastic Gradient Descent (NeurIPS 2025 · similarity 0.88)
- Statistical Inference for Gradient Boosting Regression (NeurIPS 2025 · similarity 0.87)
- Fast Last-Iterate Convergence of SGD in the Smooth Interpolation Regime (NeurIPS 2025 · similarity 0.87)
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.
