Research map / Optimization / Nonconvex and smooth optimization
On Biased Stochastic Gradient Estimation
Derek Driggs, Jingwei Liang, Carola-Bibiane Schönlieb
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 |
|---|---|---|---|
| use moderate bias theta in (1, n) in B-SAGA instead of unbiased SAGA (theta=1) (ablation) | B-SAGA · ridge regression and LASSO (australian, mushrooms, phishing, ijcnn1) | F(x_k) - F(x*): theta = 10 best among {1, 10, 100, n} | Improved |
| use small bias theta in [0.8, 1.5] in B-SVRG (ablation) | B-SVRG · ridge regression and LASSO | F(x_k) - F(x*): theta in {0.5, 0.8, 1, 1.5} | Mixed |
| use biased SARGE estimator instead of SAGA, SVRG and SARAH | SARGE · ridge regression and LASSO | F(x_k) - F(x*) | Improved |
| use SARAH instead of SVRG | SARAH · ridge regression and LASSO | F(x_k) - F(x*) | No clear effect |
| use smaller theta (more bias toward below 1) in B-SAGA and B-SVRG (ablation) | B-SAGA, B-SVRG · non-negative PCA (non-convex) | F(x_k) - F(x*) | Improved |
| use SARAH and SARGE instead of SAGA and SVRG | SARAH, SARGE · non-negative PCA (non-convex) | F(x_k) - F(x*) | No clear effect |
Closest papers by meaning
- Stochastic Bias-Reduced Gradient Methods (NeurIPS 2021 · similarity 0.92)
- Generalization Analysis of Biased Stochastic Gradient Methods for Minimax Problems (NeurIPS 2026 · similarity 0.89)
- Universal Gradient Methods for Stochastic Convex Optimization (ICML 2024 · similarity 0.89)
- The Sample Complexity of Gradient Descent in Stochastic Convex Optimization (NeurIPS 2024 · similarity 0.88)
- On the Optimal Construction of Unbiased Gradient Estimators for Zeroth-Order Optimization (NeurIPS 2025 · similarity 0.88)
- A Study of First-Order Methods with a Deterministic Relative-Error Gradient Oracle (ICML 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.
