Research map / Learning theory and algorithms / Matrix and tensor methods
Improved Subsampled Randomized Hadamard Transform for Linear SVM
Zijian Lei, Liang Lan
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 |
|---|---|---|---|
| replace uniform sampling with supervised non-uniform sampling | SRHT · rcv1-binary | test accuracy (%): 79.08 -> 87.81 | Improved |
| replace uniform sampling with supervised non-uniform sampling | SRHT · news20-binary | test accuracy (%): 70.58 -> 78.33 | Improved |
| replace uniform sampling with supervised non-uniform sampling | SRHT · real-sim | test accuracy (%): 72.80 -> 80.04 | Improved |
| replace uniform sampling with supervised non-uniform sampling | SRHT · mushrooms | test accuracy (%): 92.45 -> 96.25 | Improved |
| replace uniform sampling with deterministic top-r sampling | SRHT · rcv1-binary | test accuracy (%): 79.08 -> 82.81 | Improved |
| replace uniform sampling with deterministic top-r sampling | SRHT · gisette | test accuracy (%): 91.77 -> 93.20 | Improved |
| replace uniform sampling with norm-proportional sampling | SRHT · gisette | test accuracy (%): 91.77 -> 91.17 | Hurt |
| replace uniform sampling with norm-proportional sampling | SRHT · news20-binary | test accuracy (%): 70.58 -> 70.36 | No clear effect |
| replace uniform sampling with norm-proportional sampling | SRHT · real-sim | test accuracy (%): 72.80 -> 73.04 | No clear effect |
| use deterministic top-r sampling instead of importance sampling (nps) | ISRHT · real-sim | test accuracy (%): 73.04 -> 76.81 | Improved |
| use ISRHT-supervised instead of PCA | PCA · gisette | test accuracy (%): 96.50 -> 93.90 | Hurt |
| use ISRHT-supervised instead of supervised proposal on time | SRHT · real-sim | training time (s): 23.3s -> 44.3s | Hurt |
Closest papers by meaning
- Block Subsampled Randomized Hadamard Transform for Nyström Approximation on Distributed Architectures (ICML 2023 · similarity 0.88)
- Random features for Grassmannian kernel approximation with bounded rank-one projections (NeurIPS 2026 · similarity 0.84)
- Fundamental Bias in Inverting Random Sampling Matrices with Application to Sub-sampled Newton (ICML 2025 · similarity 0.84)
- Debiasing Random Oblique Projections for Subsampled OLS and Fast CUR in High Dimensions (NeurIPS 2026 · similarity 0.82)
- Sliced Wasserstein with Random-Path Projecting Directions (ICML 2024 · similarity 0.82)
- Optimization Can Learn Johnson Lindenstrauss Embeddings (NeurIPS 2024 · similarity 0.82)
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.
