Research map / Optimization / Nonconvex and smooth optimization
Improved high-dimensional estimation with Langevin dynamics and stochastic weight averaging
Stanley Wei, Alex Damian, Jason D. Lee
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 |
|---|---|---|---|
| average the Langevin iterates over time instead of using the last iterate (odd k*) (ablation) | Langevin dynamics on the sphere · single-index model simulation (d=100, k*=3,5, Hermite link) | recovery of planted direction | Improved |
| use the first-order time-averaged estimator when k* is even (ablation) | Langevin dynamics with averaging · single-index model simulation (d=100, k*=4) | recovery of planted direction | Hurt |
| use top eigenvector of time-averaged theta theta^T when k* is even (ablation) | Langevin dynamics with averaging · single-index model simulation (d=100, k*=4) | recovery of planted direction | Improved |
Closest papers by meaning
- Smoothing the Landscape Boosts the Signal for SGD: Optimal Sample Complexity for Learning Single Index Models (NeurIPS 2023 · similarity 0.90)
- Time-independent Generalization Bounds for SGLD in Non-convex Settings (NeurIPS 2021 · similarity 0.89)
- Generalization of noisy SGD in unbounded non-convex settings (ICML 2025 · similarity 0.89)
- Online stochastic gradient descent on non-convex losses from high-dimensional inference (JMLR 2021 · similarity 0.88)
- Nonasymptotic Analysis of Stochastic Gradient Descent with the Richardson–Romberg Extrapolation (ICLR 2025 · similarity 0.88)
- High-dimensional Limit of SGD for Diagonal Linear Networks (NeurIPS 2026 · 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.
