Research map / Optimization / Optimizers
Stochastic Adaptive Quasi-Newton Methods for Minimizing Expected Values
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 |
|---|---|---|---|
| Replace fixed step size with adaptive step size t_k = alpha_k/(1+alpha_k delta_k) in SA-GD | SA-GD · penalized least squares random design (p=100, 500; rho varied) | log(f(x)-f(x*)) vs CPU time | Mixed |
| Use increasing sample sizes m_k = 1/2 p + 1.01^k in SA-GD (ablation) | SA-GD · penalized least squares random design (p=100, 500) | fluctuation/descent of log optimality gap vs CPU time | No clear effect |
| Use increasing sample sizes m_k = 1/2 p + 1.01^k in SA-BFGS (ablation) | SA-BFGS · penalized least squares random design (p=100, 500) | log optimality gap vs CPU time | Mixed |
| Fall back to SA-GD step when Wolfe condition fails in SA-BFGS (SA-BFGS-GD) (ablation) | SA-BFGS · penalized least squares random design (p=100) | log optimality gap vs CPU time | Helped |
| Use fixed sample size m_k=p with BFGS instead of GD direction (ablation) | SA-BFGS vs SA-GD · penalized least squares random design | log optimality gap vs CPU time | Hurt |
Closest papers by meaning
- Greedy and Random Quasi-Newton Methods with Faster Explicit Superlinear Convergence (NeurIPS 2021 · similarity 0.90)
- Exploiting Local Convergence of Quasi-Newton Methods Globally: Adaptive Sample Size Approach (NeurIPS 2021 · similarity 0.90)
- An Online Adaptive Sampling Algorithm for Stochastic Difference-of-convex Optimization with Time-varying Distributions (ICML 2025 · similarity 0.88)
- Incremental Quasi-Newton Methods with Faster Superlinear Convergence Rates (AAAI 2024 · similarity 0.87)
- Revisiting the Last-Iterate Convergence of Stochastic Gradient Methods (ICLR 2024 · similarity 0.87)
- Stochastic Zeroth-Order Optimization under Strongly Convexity and Lipschitz Hessian: Minimax Sample Complexity (NeurIPS 2024 · 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.