Research map / Optimization / Optimizers
Large Scale Online Kernel Learning
Jing Lu, Steven C.H. Hoi, Jialei Wang, Peilin Zhao, Zhi-Yong Liu
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 Nystrom approximation (NOGD) instead of random-removal budget (BSGD-R) | budget SGD · codrna batch | accuracy (%): 67.16 -> 95.92 | Improved |
| use NOGD instead of BSGD-M merging | budget SGD · codrna batch | train time (s): 132.50 -> 2.55 | Improved |
| use NOGD instead of exact LIBSVM | LIBSVM · KDD99 batch | train time (s): 20772.10 -> 20.71 | Improved |
| use FOGD instead of kernel OGD | OGD · w7a online binary | mistake rate (%): 2.96 -> 2.75 | Improved |
| use FOGD instead of BOGD | BOGD · a9a online binary | mistake rate (%): 27.9 -> 17.4 | Improved |
| use FOGD instead of kernel OGD | OGD · a9a online binary | time (s): 1549.5 -> 1.8 | Improved |
| use NOGD instead of BOGD | BOGD · acoustic multi-class | mistake rate (%): 58.0 -> 37.8 | Improved |
| use NOGD instead of FOGD at small budget B=100 | FOGD · german online binary | mistake rate (%): 29.9 -> 30.4 | No clear effect |
| use NOGD instead of FOGD at large budget in batch | FOGD · codrna batch | accuracy (%): 94.24 -> 95.92 | Improved |
| increase rho_f or rho_n (more Fourier components or rank) (ablation) | FOGD and NOGD · multi-class datasets (B=200) | mistake rate | Mixed |
| use Projectron++ instead of BOGD (projection vs removal) | BOGD · usps multi-class | mistake rate (%): 23.2 -> 9.9 | Improved |
| use BPA-S in multi-class instead of binary | BPA-S · multi-class | time cost | Hurt |
Closest papers by meaning
- Parsimonious Online Learning with Kernels via Sparse Projections in Function Space (JMLR 2019 · similarity 0.88)
- Efficient online learning with kernels for adversarial large scale problems (NeurIPS 2019 · similarity 0.88)
- Dual Space Gradient Descent for Online Learning (NeurIPS 2016 · similarity 0.88)
- Limited Memory Online Gradient Descent for Kernelized Pairwise Learning with Dynamic Averaging (AAAI 2024 · similarity 0.87)
- Fast Training of Large Kernel Models with Delayed Projections (NeurIPS 2025 · similarity 0.86)
- Joker: Joint Optimization Framework for Lightweight Kernel Machines (ICML 2025 · similarity 0.86)
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.
