Research map / Learning theory and algorithms / Kernels and regression theory
Parsimonious Online Learning with Kernels via Sparse Projections in Function Space
Alec Koppel, Garrett Warnell, Ethan Stump, Alejandro Ribeiro
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 POLK sparse projection instead of fixed-budget BSGD at the same model order | Multi-KSVM · multidist | test error (%): 21.8 -> 3.98 | Improved |
| use POLK instead of BSGD at the same model order | Multi-KSVM · mnist | test error (%): 2.67 -> 2.46 | Improved |
| use POLK instead of BSGD at the same model order | Multi-KSVM · brodatz | test error (%): 4.72 -> 4.53 | Mixed |
| use POLK instead of batch LIBSVM | Multi-KSVM · multidist | model points: 656 -> 16 | Improved |
| use POLK instead of batch LIBSVM | Multi-KSVM · mnist | test error (%): 1.50 -> 2.46 | Hurt |
| use POLK instead of batch LIBSVM | Multi-KSVM · mnist | model points: 16118 -> 1086 | Improved |
| use POLK instead of dense L-BFGS | Multi-KSVM · multidist | test error (%): 4.08 -> 3.98 | Improved |
| use POLK instead of L-BFGS | Multi-Logistic · brodatz | test error (%): 4.00 -> 4.41 | Hurt |
| use POLK instead of L-BFGS | Multi-Logistic · brodatz | model points: 10000 -> 1833 | Improved |
| use POLK instead of import vector machine (IVM) | Multi-Logistic · multidist | test error (%): 4.08 -> 4.36 | No clear effect |
| reduce parsimony constant K (larger model order) (ablation) | POLK Multi-Logistic · multidist | test error (%): 4%-5% error, model order 75 to 16 | Mixed |
| reduce parsimony constant K (larger model order) (ablation) | POLK Multi-KSVM · multidist | test error (%): 16 -> 129 model order | Improved |
Closest papers by meaning
- Large Scale Online Kernel Learning (JMLR 2016 · similarity 0.88)
- Efficient Second-Order Online Kernel Learning with Adaptive Embedding (NeurIPS 2017 · similarity 0.86)
- Hierarchically Compositional Kernels for Scalable Nonparametric Learning (JMLR 2017 · similarity 0.86)
- Dual Space Gradient Descent for Online Learning (NeurIPS 2016 · similarity 0.86)
- Limited Memory Online Gradient Descent for Kernelized Pairwise Learning with Dynamic Averaging (AAAI 2024 · similarity 0.85)
- Bayesian Optimization on Function Spaces via Sparse RKHS Manifolds (NeurIPS 2026 · 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.
