Research map / Learning theory and algorithms / Kernels and regression theory
Online Multi-Kernel Learning with Graph-Structured Feedback
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 max kernels per selective node M (e.g. 1 to 7 at J=1) (ablation) | OMKL-GF · Air Quality, Istanbul, Twitter, Tom's Hardware | MSE: Air Quality 8.1→4.2; Istanbul 61.9→13.3; Twitter 16.5→5.3; Tom's 12.3→3.7 | Helped |
| Reduce number of selective nodes J (fewer, e.g. J=1 vs J=2 or 4) (ablation) | OMKL-GF · Air Quality, Istanbul, Twitter, Tom's Hardware, Naval | MSE: Air Quality M=1: J=1 8.1 vs J=4 11.2; Naval M=1: J=1 20.1 vs J=3 18.6 (reverse) | Helped |
| Increase number of selective nodes J from 1 to 3 (Naval Propulsion) (ablation) | OMKL-GF · Naval Propulsion Plants | MSE: M=1: 20.1 (J=1) vs 18.6 (J=3); M=10: 1.4 (J=1) vs 1.7 (J=3) | Mixed |
| Use graph-structured kernel selection (OMKL-GF, M=7-10, J=1) instead of all-kernel Raker | OMKL-GF vs Raker · Air Quality, Twitter, Tom's Hardware | MSE: Air Quality 3.9 vs 4.7; Twitter 4.7 vs 5.3; Tom's 3.5 vs 4.6 | Helped |
| Graph-structured kernel selection vs Raker on Istanbul and Naval datasets | OMKL-GF vs Raker · Istanbul Stock Exchange, Naval Propulsion Plants | MSE: Istanbul 11.3 vs 11.3 (M=17,J=1); Naval 1.4 vs 1.4 (M=10,J=1) | No clear effect |
| Graph-structured kernel selection vs OMKR-FA (random feature MKL with all kernels) | OMKL-GF vs OMKR-FA · all five datasets (M=1, J=1 and larger M) | MSE: Air Quality 8.1 vs 41.2 (M=1); Naval 1.4 vs 294.9 (M=10) | Helped |
| Graph-structured kernel selection vs exact OMKR (no random features) | OMKL-GF vs OMKR · all datasets | MSE: OMKR 3.3 vs OMKL-GF best 3.9 (Air Quality); OMKR 1.0 vs 1.4 (Naval) | Hurt |
| Graph-structured selection reduces execution time vs Raker | OMKL-GF vs Raker · Naval Propulsion Plants | execution time (s): 1.24s (M=1,J=1) / 2.49s (M=10,J=1) vs 4.14s Raker | Helped |
| Lazy graph generation (skip graph once squared error < 1e-4) | OMKL-GF · all datasets | MSE: not separately reported | No clear effect |
Closest papers by meaning
- Regret Bounds for Online Kernel Selection in Continuous Kernel Space (AAAI 2021 · similarity 0.87)
- Efficient Second-Order Online Kernel Learning with Adaptive Embedding (NeurIPS 2017 · similarity 0.87)
- Personalized Online Federated Learning with Multiple Kernels (NeurIPS 2022 · similarity 0.86)
- Multi-View Randomized Kernel Classification via Nonconvex Optimization (AAAI 2024 · similarity 0.86)
- On the Target-kernel Alignment: a Unified Analysis with Kernel Complexity (NeurIPS 2024 · similarity 0.84)
- Joker: Joint Optimization Framework for Lightweight Kernel Machines (ICML 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.