Research map / Learning theory and algorithms / Kernels and regression theory
Probabilistic Kernel Function for Fast Angle Testing
Kejing Lu, Chuan Xiao, Yoshiharu Ishikawa
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 Gaussian projection vectors (CEOs) with S_pol structured projections | KS1 · Word k-MIPS, Probe@10 | recall (%): 34.106 -> 34.395 | Improved |
| replace Gaussian projection vectors (CEOs) with S_pol structured projections | KS1 · Word k-MIPS, Probe@10K | recall (%): 98.182 -> 98.404 | Improved |
| replace Gaussian projection vectors (CEOs) with S_pol structured projections | KS1 · GloVe1M k-MIPS, Probe@10K | recall (%): 63.545 -> 64.355 | Improved |
| replace Gaussian projection vectors (CEOs) with S_pol structured projections | KS1 · GloVe2M k-MIPS, Probe@10K | recall (%): 54.916 -> 55.394 | Improved |
| replace Gaussian projection vectors (CEOs) with S_pol structured projections | KS1 · GloVe2M k-MIPS, Probe@10 | recall (%): 2.070 -> 1.996 | Hurt |
| replace Gaussian projection vectors (CEOs) with S_sym structured projections | KS1 · GloVe2M k-MIPS, Probe@10 | recall (%): 2.070 -> 2.064 | Hurt |
| replace Gaussian projection vectors (CEOs) with S_sym structured projections | KS1 · GloVe1M k-MIPS, Probe@10K | recall (%): 63.545 -> 64.041 | Improved |
| use S_pol configuration instead of S_sym (ablation) | KS1 · GloVe1M k-MIPS, Probe@10K | recall (%): 64.041 -> 64.355 | Improved |
| use S_pol configuration instead of S_sym (ablation) | KS1 · Word k-MIPS, Probe@1K | recall (%): 90.265 -> 90.678 | Improved |
| use S_pol configuration instead of S_sym (ablation) | KS1 · GloVe2M k-MIPS, Probe@10 | recall (%): 2.064 -> 1.996 | Hurt |
| add KS2 routing test to HNSW | HNSW · ANNS (datasets except Word), recall-QPS | QPS: factor of 2.5 to 3 | Improved |
| replace PEOs test with KS2 test in HNSW | HNSW+PEOs · ANNS (datasets except Word), recall-QPS | QPS: 1.1 to 1.3 times faster | Improved |
Closest papers by meaning
- Nonparametric Testing Under Randomized Sketching (TPAMI 2021 · similarity 0.85)
- The GeometricKernels Package: Heat and Matérn Kernels for Geometric Learning on Manifolds, Meshes, and Graphs (JMLR 2025 · similarity 0.84)
- Omnipresent Yet Overlooked: Heat Kernels in Combinatorial Bayesian Optimization (NeurIPS 2025 · similarity 0.83)
- Random features for Grassmannian kernel approximation with bounded rank-one projections (NeurIPS 2026 · similarity 0.82)
- Kernel conditional tests from learning-theoretic bounds (NeurIPS 2025 · similarity 0.82)
- Probabilistic Routing for Graph-Based Approximate Nearest Neighbor Search (ICML 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.
