Research map / Optimization / Nonconvex and smooth optimization
Online to Offline Conversions, Universality and Adaptive Minibatch Sizes
Kfir Levy
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 SC-AdaNGD variants with k in {1, 1.1, 2} instead of GD or line-search | Strongly convex smooth quadratic R · Quadratic R | objective loss vs. gradient-oracle calls | Improved |
| Use accelerated GD rather than SC-AdaNGD | Strongly convex smooth quadratic R · Quadratic R | objective loss vs. gradient-oracle calls | Improved |
| Set SC-AdaNGD k=1.1 rather than GD | Strongly convex smooth quadratic R · Quadratic R | objective loss vs. gradient-oracle calls | Improved |
| Use SC-AdaNGD k=2 rather than the tested GD variants | Strongly convex nonsmooth F · Nonsmooth F | objective loss vs. gradient-oracle calls | Improved |
| Use a constant learning rate rather than SC-AdaNGD | GD on strongly convex nonsmooth F · Nonsmooth F | objective loss vs. gradient-oracle calls | Hurt |
| Use decaying-rate GD rather than SC-AdaNGD k=2 | GD on strongly convex nonsmooth F · Nonsmooth F | objective loss vs. gradient-oracle calls | Hurt |
| Use the AE sample count n_s as the inverse-squared-gradient estimate (ablation) | LazySGD · Stochastic optimization problems | performance | Hurt |
| Use 1/||g̃_s||² instead of AE sample count n_s for the minibatch rule (ablation) | LazySGD · Stochastic optimization problems | performance | Improved |
| Use modified LazySGD minibatches rather than fixed-size minibatch SGD | LazySGD · Several simple stochastic problems | performance | No clear effect |
Closest papers by meaning
- Optimistic Online-to-Batch Conversions for Accelerated Convergence and Universality (NeurIPS 2025 · similarity 0.90)
- Gradient-Variation Online Adaptivity for Accelerated Optimization with Hölder Smoothness (NeurIPS 2025 · similarity 0.90)
- Anytime Online-to-Batch, Optimism and Acceleration (ICML 2019 · similarity 0.90)
- An Online Adaptive Sampling Algorithm for Stochastic Difference-of-convex Optimization with Time-varying Distributions (ICML 2025 · similarity 0.88)
- Gradient-Variation Online Learning under Generalized Smoothness (NeurIPS 2024 · similarity 0.87)
- Faster Stochastic Optimization with Arbitrary Delays via Adaptive Asynchronous Mini-Batching (ICML 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.
