Research map / Learning theory and algorithms / Matrix and tensor methods
Randomized Clustered Nystrom for Large-Scale Kernel Machines
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 |
|---|---|---|---|
| Use QR-based rank-r approximation instead of restricting rank through the inner matrix W | Nyström approximation · 3×3 toy kernel matrix | normalized Frobenius kernel approximation error: 0.99 -> 0.01 | Helped |
| Select landmarks using K-means on randomized low-dimensional projections rather than uniform sampling | Nyström kernel approximation · dna and protein | normalized Frobenius kernel approximation error | Helped |
| Select landmarks using K-means on randomized low-dimensional projections rather than uniform sampling | Nyström kernel approximation · mnist and epsilon | normalized Frobenius kernel approximation error | Helped |
| Select landmarks using K-means on randomized low-dimensional projections rather than original-space K-means | Nyström kernel approximation · mnist and epsilon | runtime: an order of magnitude | Helped |
| Increase compression factor γ (ablation) | Randomized Clustered Nyström · dna and protein | normalized Frobenius kernel approximation error | Helped |
| Increase compression factor γ (ablation) | Randomized Clustered Nyström · dna and protein | memory and computation savings versus Clustered Nyström: 1/γ | Hurt |
| Use γ=0.2 rather than Clustered Nyström | Randomized Clustered Nyström · protein | normalized Frobenius kernel approximation error | Helped |
| Use more than r landmarks and increase m (ablation) | Nyström via QR Decomposition · satimage | normalized Frobenius kernel approximation error | Helped |
| Increase the number of landmarks m (ablation) | Standard Nyström · satimage | normalized Frobenius kernel approximation error | Mixed |
| Set the number of landmarks to m=r | Randomized Clustered Nyström · epsilon | normalized Frobenius kernel approximation error: m=r versus m=10r | Helped |
| Use Nyström via QR Decomposition rather than standard Nyström | Nyström approximation · synthetic data set | normalized Frobenius kernel approximation error: m=4 versus m=40 | Helped |
Closest papers by meaning
- Nyström Method with Kernel K-means++ Samples as Landmarks (ICML 2017 · similarity 0.91)
- Computationally Efficient Nyström Approximation using Fast Transforms (ICML 2016 · similarity 0.89)
- Incremental Nyström-based Multiple Kernel Clustering (AAAI 2025 · similarity 0.88)
- Recursive Sampling for the Nystrom Method (NeurIPS 2017 · similarity 0.88)
- Random features for Grassmannian kernel approximation with bounded rank-one projections (NeurIPS 2026 · similarity 0.85)
- Scalable Multiple Kernel Clustering: Learning Clustering Structure from Expectation (ICML 2024 · similarity 0.85)
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.