Research map / Optimization / Optimizers
Dual Space Gradient Descent for Online Learning
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 merging size k from 1 to 150 (ablation) | DualSGD-Hinge · cod-rna | mistake rate | Hurt |
| increase merging size k from 1 to 150 (ablation) | DualSGD-Hinge · cod-rna | running time | Helped |
| increase random feature dimension D from 100 to 800 at B = 50 (ablation) | DualSGD-Hinge · cod-rna | mistake rate: 25% | Helped |
| increase random feature dimension D from 100 to 800 at B = 50 (ablation) | DualSGD-Hinge · cod-rna | wall-clock time: 76% | Hurt |
| increase budget size B from 50 to 400 at D = 100 (ablation) | DualSGD-Hinge · cod-rna | mistake rate: 1.5% | Helped |
| increase budget size B from 50 to 400 at D = 100 (ablation) | DualSGD-Hinge · cod-rna | wall-clock time: 54% | Hurt |
| replace Hinge loss with Logistic loss (ablation) | DualSGD · cod-rna | mistake rate: 4.92 -> 4.83 | No clear effect |
| replace Hinge loss with Logistic loss (ablation) | DualSGD · ijcnn1 | mistake rate: 8.35 -> 8.82 | No clear effect |
| replace epsilon-insensitive loss with l1 loss (ablation) | DualSGD · year | RMSE: 0.13 -> 0.12 | Helped |
| replace budget merging (BSGD-M) with k-merging plus random-feature provision vector | BSGD-M · cod-rna | mistake rate: 5.33 -> 4.92 | Helped |
| replace budget merging (BSGD-M) with k-merging plus random-feature provision vector | BSGD-M · cod-rna | execution time (s): 184.58 -> 28.29 | Helped |
| add provision vector in random-feature space to FOGD-style random features | FOGD · cod-rna | mistake rate: 7.15 -> 4.92 | Helped |
Closest papers by meaning
- Limited Memory Online Gradient Descent for Kernelized Pairwise Learning with Dynamic Averaging (AAAI 2024 · similarity 0.87)
- Efficient online learning with kernels for adversarial large scale problems (NeurIPS 2019 · similarity 0.87)
- Joker: Joint Optimization Framework for Lightweight Kernel Machines (ICML 2025 · similarity 0.86)
- Learning Curves of Stochastic Gradient Descent in Kernel Regression (ICML 2025 · similarity 0.85)
- Random features for Grassmannian kernel approximation with bounded rank-one projections (NeurIPS 2026 · similarity 0.84)
- Fast Training of Large Kernel Models with Delayed Projections (NeurIPS 2025 · similarity 0.84)
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.