Research map / Learning theory and algorithms / Clustering
Speeding up k-means by approximating Euclidean distances via block vectors
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 |
|---|---|---|---|
| Add block-vector lower bounds to yinyang k-means (fast yinyang) | yinyang · 9 datasets | speed-up: up to 2.4 | Helped |
| Yinyang outperforms fast yinyang in remaining cases | fast yinyang · 9 datasets | speed-up: 1.12 | Hurt |
| Increase number of clusters k | kmeans optimized · usps | avoided distance calculations: 0.956 -> 0.976 | Helped |
| Increase number of clusters k | kmeans optimized · sector | avoided distance calculations: 0.336 -> 0.739 | Helped |
| Increase number of clusters k | kmeans optimized · real-sim | avoided distance calculations: 0.035 -> 0.237 | Helped |
| Increase number of clusters k | kmeans optimized · kdd2010 | avoided distance calculations: 0.987 -> 0.999 | Helped |
| Increase k from 100 to 10000 | fast yinyang · mnist800k | speed-up: 23.6 -> 36.5 | Helped |
| Increase k from 100 to 1000 to 10000 | fast yinyang · caltech101 | speed-up: 14.6 -> 27.4 -> 15.8 | Mixed |
| Use kmeans optimized instead of yinyang at highest k | kmeans optimized · 9 datasets, highest k | speed-up | Helped |
| Block vector lower bound never avoids a calculation | kmeans optimized · runtime | runtime overhead: 30% | Hurt |
| Block vector lower bound always avoids full distance calculation | kmeans optimized · runtime | runtime reduction: 70% | Helped |
| Use kmeans optimized instead of yinyang on usps at k=1000 | kmeans optimized · usps | speed-up: 17.5 -> 7.6 | Hurt |
Closest papers by meaning
- A Sampling-Based Approach for Efficient Clustering in Large Datasets (CVPR 2022 · similarity 0.87)
- Fast k-means with accurate bounds (ICML 2016 · similarity 0.87)
- Clustering Stable Instances of Euclidean k-means. (NeurIPS 2017 · similarity 0.86)
- Improved Guarantees for k-means++ and k-means++ Parallel (NeurIPS 2020 · similarity 0.86)
- An Objective for Hierarchical Clustering in Euclidean Space and Its Connection to Bisecting K-means (AAAI 2020 · similarity 0.86)
- Near-Linear Time Approximation Algorithms for k-means with Outliers (ICML 2024 · similarity 0.86)
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.