Research map / Optimization / Nonconvex and smooth optimization
A Simple Proximal Stochastic Gradient Method for Nonsmooth Nonconvex Optimization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 ProxSVRG+ with minibatch size $b=1/\epsilon^{2/3}$ instead of ProxGD | Nonsmooth nonconvex finite-sum optimization · Nonsmooth nonconvex finite-sum | SFO calls: $O(n/\epsilon)$ -> $O\big(\frac{n}{\epsilon^{2/3}} + \frac{1}{\epsilon^{5/3}}\big)$ | Helped |
| Use ProxSVRG+ with minibatch size $b=1/\epsilon^{2/3}$ under bounded variance and $n>1/\epsilon$ | Nonsmooth nonconvex finite-sum optimization · Nonsmooth nonconvex finite-sum | SFO calls: $O(n/\epsilon)$ -> $O\big(\frac{1}{\epsilon^{5/3}}\big)$ | Helped |
| Use ProxSVRG+ with minibatch size $b=n^{2/3}$ | Nonsmooth nonconvex finite-sum optimization · Nonsmooth nonconvex finite-sum | SFO calls: $O\big(\frac{n^{2/3}}{\epsilon}\big)$ | No clear effect |
| Use ProxSVRG+ with minibatch size $b=1$ or $b=n$ | ProxSVRG+ · Nonsmooth nonconvex finite-sum | SFO calls: $O(n/\epsilon)$ | No clear effect |
| Use ProxSVRG+ with moderate minibatch sizes rather than ProxSVRG's larger preferred minibatch | NN-PCA on MNIST and a9a · MNIST and a9a | SFO calls to reach the reported performance | Helped |
| Reduce ProxSVRG+ minibatch size to $b=64$ | NN-PCA on MNIST and a9a · MNIST and a9a | performance | No clear effect |
| Use ProxSVRG+ with $b<n^{2/3}$ instead of ProxSVRG | NN-PCA on MNIST and a9a · MNIST and a9a | PO calls | Helped |
| Use ProxSVRG+ under the PL condition without restarting | Nonsmooth nonconvex finite-sum optimization · PL-condition problems | PO calls: $O(\frac{1}{\mu}\log\frac{1}{\epsilon})$ | Helped |
Closest papers by meaning
- Proximal Stochastic Methods for Nonsmooth Nonconvex Finite-Sum Optimization (NeurIPS 2016 · similarity 0.95)
- Faster Gradient-Free Proximal Stochastic Methods for Nonconvex Nonsmooth Optimization (AAAI 2019 · similarity 0.93)
- Convergence Analysis of Proximal Gradient with Momentum for Nonconvex Optimization (ICML 2017 · similarity 0.91)
- A Bregman Proximal Stochastic Gradient Method with Extrapolation for Nonconvex Nonsmooth Problems (AAAI 2024 · similarity 0.91)
- Adaptive Proximal Gradient Method for Convex Optimization (NeurIPS 2024 · similarity 0.90)
- Stochastic Smoothed Primal-Dual Algorithms for Nonconvex Optimization with Linear Inequality Constraints (ICML 2025 · similarity 0.89)
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.