Research map / Learning theory and algorithms / Optimal transport
S2JSD-LSH: A Locality-Sensitive Hashing Schema for Probability Distributions
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 S2JSD-LSH instead of Super-bit LSH | LSH for probability distributions · Y Ans | mAP (256 bits): 0.8963 -> 0.9323 | Helped |
| use S2JSD-LSH instead of Super-bit LSH | LSH for probability distributions · CIFAR10 | mAP (256 bits): 0.1413 -> 0.1430 | Helped |
| use S2JSD-LSH instead of Super-bit LSH | LSH for probability distributions · CIFAR100-100 | mAP (256 bits): 0.0228 -> 0.0232 | Helped |
| use S2JSD-LSH instead of Super-bit LSH at low bits | LSH for probability distributions · MNIST | mAP: 16 bits: 0.1818 -> 0.1654; 8 bits: 0.1125 -> 0.1429 | Mixed |
| use S2JSD-LSH instead of Super-bit LSH | LSH for probability distributions · Local-Patch | mAP: 256 bits: 0.3804 -> 0.3825; 128 bits: 0.3786 -> 0.3755 | Mixed |
| use S2JSD-LSH instead of p-stable l2 LSH | LSH for probability distributions · eight datasets | mAP: MNIST 256 bits: 0.1004 -> 0.3710 | Helped |
| use information-theoretic metrics (JSD/S2JSD) instead of l2 for KNN | brute-force search · CIFAR10 | p@5: 0.478 -> 0.528 | Helped |
| use new symmetric first-order approximation instead of Endres approximation of S2JSD (ablation) | S2JSD approximation · Dir(1) simulation, 100 dims | closeness to S2JSD: 98.455% of pairs closer | Helped |
| increase partition width W from 0.1 to 1.0 (ablation) | S2JSD-LSH · CIFAR100-100 and MNIST | mAP | Hurt |
Closest papers by meaning
- Locality-Sensitive Hashing for f-Divergences: Mutual Information Loss and Beyond (NeurIPS 2019 · similarity 0.91)
- Local Density Estimation in High Dimensions (ICML 2018 · similarity 0.88)
- Learning to Hash Robustly, Guaranteed (ICML 2022 · similarity 0.87)
- Approximate Nearest Neighbor Search through Modern Error-Correcting Codes (ICLR 2023 · similarity 0.87)
- Locality Sensitive Hashing for p-Exponential Kernels with Applications to Density Estimation (NeurIPS 2026 · similarity 0.86)
- Effective Comparative Prototype Hashing for Unsupervised Domain Adaptation (AAAI 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.