Research map / Optimization / Nonconvex and smooth optimization
SAGA with Arbitrary Sampling
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 |
|---|---|---|---|
| Generalize SAGA to arbitrary proper sampling using a bias-correcting random vector | SAGA · Smooth strongly convex and composite finite-sum problems | iteration complexity | Helped |
| Choose the optimal bias-correcting random vector instead of the default vector (ablation) | SAGA-AS · Fixed arbitrary sampling; smooth strongly convex problems | iteration complexity: \min_{\theta \in \Theta(S)} \beta_i = \tfrac{1}{\sum_{C:i\in C}p_C/|C|} = \tfrac{1}{p_i \mathbb{E}^i [1/ |S|]} | Helped |
| Use importance independent sampling instead of τ-nice sampling (ablation) | SAGA-AS · Smooth strongly convex finite-sum problems | iteration complexity: at most (1+1/τ) times worse; could also be n/(τ+1) times better | Mixed |
| Increase minibatch size from τ = 1 to τ = 50 (ablation) | SAGA-AS · ijcnn1 logistic regression | epochs to convergence: increased by less than 6 | No clear effect |
| Use minibatch importance sampling instead of uniform sampling (ablation) | SAGA · ijcnn1 and w8a | epochs to convergence | Helped |
Closest papers by meaning
- A Short and Unified Convergence Analysis of the SAG, SAGA, and IAG Algorithms (ICML 2026 · similarity 0.90)
- Optimal Mini-Batch and Step Sizes for SAGA (ICML 2019 · similarity 0.90)
- A Composite Randomized Incremental Gradient Method (ICML 2019 · similarity 0.88)
- Optimization Can Learn Johnson Lindenstrauss Embeddings (NeurIPS 2024 · similarity 0.85)
- Scalable Discrete Diffusion Samplers: Combinatorial Optimization and Statistical Physics (ICLR 2025 · similarity 0.85)
- Spherical Flows for Sampling Categorical Data (NeurIPS 2026 · similarity 0.85)
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.