Research map / Bandits and online learning / Online convex optimization and games
Efficient Second Order Online Learning by Sketching
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 |
|---|---|---|---|
| Increase Oja-SON sketch size from 0 to 5 (ablation) | Oja-SON · Synthetic ill-conditioned datasets | progressive error | Helped |
| Increase Oja-SON sketch size from 5 to 10 (ablation) | Oja-SON · Synthetic ill-conditioned datasets | progressive error | Helped |
| Use sketch size 10 instead of 0 (ablation) | Oja-SON · Acoustic, census, heart, ionosphere, and letter datasets | error rate | Helped |
| Use sketch size 10 instead of 0 (ablation) | Oja-SON · Real-world benchmark datasets | error rate | No clear effect |
| Apply diagonal rescaling before Oja-SON (ablation) | Oja-SON · Real-world benchmark datasets | error rate | Helped |
| Increase sketch size from 0 to 10 after diagonal rescaling (ablation) | Diagonally rescaled Oja-SON · Real-world benchmark datasets | error rate | No clear effect |
| Use Oja's sketch instead of Frequent Directions (ablation) | Sketched Online Newton · Experiments comparing sketching options | error rate | Helped |
| Use Oja-SON without a sketch (m=0) instead of AdaGrad (ablation) | Oja-SON · Synthetic ill-conditioned datasets | accuracy | Hurt |
| Use Oja-SON instead of AdaGrad without diagonal rescaling | Oja-SON · Most real-world benchmark datasets | error rate | Hurt |
Closest papers by meaning
- Second-Order Kernel Online Convex Optimization with Adaptive Sketching (ICML 2017 · similarity 0.87)
- Incremental Randomized Sketching for Online Kernel Learning (ICML 2019 · similarity 0.86)
- Non-stationary Projection-Free Online Learning with Dynamic and Adaptive Regret Guarantees (AAAI 2024 · similarity 0.84)
- From Average Sensitivity to Small-Loss Regret Bounds under Random-Order Model (NeurIPS 2026 · similarity 0.83)
- Sketching for Convex and Nonconvex Regularized Least Squares with Sharp Guarantees (ICLR 2025 · similarity 0.83)
- Efficient Non-stationary Online Learning by Wavelets with Applications to Online Distribution Shift Adaptation (ICML 2024 · similarity 0.83)
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.