Research map / Optimization / Nonconvex and smooth optimization
Generalized-Smooth Nonconvex Optimization is As Efficient As Smooth Nonconvex Optimization
Ziyi Chen, Yi Zhou, Yingbin Liang, Zhaosong Lu
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 |
|---|---|---|---|
| Normalize gradient step by ||grad||^beta with beta in [alpha,1] (beta-GD) instead of plain GD | deterministic nonconvex optimization on L*_sym(alpha) · phase retrieval synthetic experiment | convergence speed (objective vs iteration) | Improved |
| Under-normalization with beta < alpha in beta-GD | beta-GD on L*_sym(alpha) · theoretical (convex counterexample) | convergence (divergence) | Hurt |
| Add SPIDER-style variance reduction with normalized updates (SPIDER) for expected generalized-smooth stochastic problems | stochastic nonconvex optimization on EL*_sym(alpha) · theory: sample complexity | sample complexity O(eps^-3) | Improved |
| Use SPIDER (variance reduction + normalization) instead of SGD-type methods (SGD, normalized SGD, momentum, clipped SGD) (ablation) | stochastic phase retrieval · phase retrieval stochastic experiment | final objective vs sample complexity (qualitative) | Improved |
| Use SPIDER instead of SGD-type methods (ablation) | stochastic DRO · DRO experiment | final objective vs sample complexity (qualitative) | Improved |
| Normalized GD (beta=1) vs standard GD on DRO (ablation) | deterministic DRO objective · DRO experiment | objective vs iteration (qualitative) | Improved |
Closest papers by meaning
- Toward a Unified Theory of Gradient Descent under Generalized Smoothness (ICML 2025 · similarity 0.90)
- Nonlinearly Preconditioned Gradient Methods under Generalized Smoothness (ICML 2025 · similarity 0.88)
- Optimizing $(L_0, L_1)$-Smooth Functions by Gradient Methods (ICLR 2025 · similarity 0.88)
- Efficiently Escaping Saddle Points under Generalized Smoothness via Self-Bounding Regularity (NeurIPS 2025 · similarity 0.88)
- Methods for Convex $(L_0,L_1)$-Smooth Optimization: Clipping, Acceleration, and Adaptivity (ICLR 2025 · similarity 0.88)
- Escaping saddle points without Lipschitz smoothness: the power of nonlinear preconditioning (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.
