Model
LSH banding: the S-curve
Pairs above the threshold are found; pairs below it are almost never compared.
- Similarity
- Jaccard similarity of the pair
- Bands
- bands
- Rows per band
- rows (hashes) per band
How it behaves
One input on the horizontal axis, the others held fixed.
b = 16, r = 8 (threshold ≈ 0.71)b = 32, r = 4 (threshold ≈ 0.42)
Data table
| Jaccard similarity J | series | P(candidate) |
|---|---|---|
| 0 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.013 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.025 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.038 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.05 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.063 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.075 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.088 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.1 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.113 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.125 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.138 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.15 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.163 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.175 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.188 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.2 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.212 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.225 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.237 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.25 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.263 | b = 16, r = 8 (threshold ≈ 0.71) | 0 |
| 0.275 | b = 16, r = 8 (threshold ≈ 0.71) | 0.001 |
| 0.287 | b = 16, r = 8 (threshold ≈ 0.71) | 0.001 |
| 0.3 | b = 16, r = 8 (threshold ≈ 0.71) | 0.001 |
| 0.313 | b = 16, r = 8 (threshold ≈ 0.71) | 0.001 |
| 0.325 | b = 16, r = 8 (threshold ≈ 0.71) | 0.002 |
| 0.338 | b = 16, r = 8 (threshold ≈ 0.71) | 0.003 |
| 0.35 | b = 16, r = 8 (threshold ≈ 0.71) | 0.004 |
| 0.362 | b = 16, r = 8 (threshold ≈ 0.71) | 0.005 |
| 0.375 | b = 16, r = 8 (threshold ≈ 0.71) | 0.006 |
| 0.388 | b = 16, r = 8 (threshold ≈ 0.71) | 0.008 |
| 0.4 | b = 16, r = 8 (threshold ≈ 0.71) | 0.01 |
| 0.412 | b = 16, r = 8 (threshold ≈ 0.71) | 0.013 |
| 0.425 | b = 16, r = 8 (threshold ≈ 0.71) | 0.017 |
| 0.438 | b = 16, r = 8 (threshold ≈ 0.71) | 0.021 |
| 0.45 | b = 16, r = 8 (threshold ≈ 0.71) | 0.027 |
| 0.463 | b = 16, r = 8 (threshold ≈ 0.71) | 0.033 |
| 0.475 | b = 16, r = 8 (threshold ≈ 0.71) | 0.041 |
| 0.487 | b = 16, r = 8 (threshold ≈ 0.71) | 0.05 |
| 0.5 | b = 16, r = 8 (threshold ≈ 0.71) | 0.061 |
| 0.512 | b = 16, r = 8 (threshold ≈ 0.71) | 0.074 |
| 0.525 | b = 16, r = 8 (threshold ≈ 0.71) | 0.088 |
| 0.537 | b = 16, r = 8 (threshold ≈ 0.71) | 0.106 |
| 0.55 | b = 16, r = 8 (threshold ≈ 0.71) | 0.126 |
| 0.563 | b = 16, r = 8 (threshold ≈ 0.71) | 0.149 |
| 0.575 | b = 16, r = 8 (threshold ≈ 0.71) | 0.175 |
| 0.588 | b = 16, r = 8 (threshold ≈ 0.71) | 0.204 |
| 0.6 | b = 16, r = 8 (threshold ≈ 0.71) | 0.237 |
| 0.613 | b = 16, r = 8 (threshold ≈ 0.71) | 0.274 |
| 0.625 | b = 16, r = 8 (threshold ≈ 0.71) | 0.314 |
| 0.637 | b = 16, r = 8 (threshold ≈ 0.71) | 0.358 |
| 0.65 | b = 16, r = 8 (threshold ≈ 0.71) | 0.404 |
| 0.662 | b = 16, r = 8 (threshold ≈ 0.71) | 0.454 |
| 0.675 | b = 16, r = 8 (threshold ≈ 0.71) | 0.506 |
| 0.688 | b = 16, r = 8 (threshold ≈ 0.71) | 0.559 |
| 0.7 | b = 16, r = 8 (threshold ≈ 0.71) | 0.613 |
| 0.713 | b = 16, r = 8 (threshold ≈ 0.71) | 0.667 |
| 0.725 | b = 16, r = 8 (threshold ≈ 0.71) | 0.719 |
| 0.738 | b = 16, r = 8 (threshold ≈ 0.71) | 0.769 |
| 0.75 | b = 16, r = 8 (threshold ≈ 0.71) | 0.815 |
| 0.762 | b = 16, r = 8 (threshold ≈ 0.71) | 0.857 |
| 0.775 | b = 16, r = 8 (threshold ≈ 0.71) | 0.893 |
| 0.787 | b = 16, r = 8 (threshold ≈ 0.71) | 0.923 |
| 0.8 | b = 16, r = 8 (threshold ≈ 0.71) | 0.947 |
| 0.813 | b = 16, r = 8 (threshold ≈ 0.71) | 0.966 |
| 0.825 | b = 16, r = 8 (threshold ≈ 0.71) | 0.979 |
| 0.838 | b = 16, r = 8 (threshold ≈ 0.71) | 0.988 |
| 0.85 | b = 16, r = 8 (threshold ≈ 0.71) | 0.994 |
| 0.863 | b = 16, r = 8 (threshold ≈ 0.71) | 0.997 |
| 0.875 | b = 16, r = 8 (threshold ≈ 0.71) | 0.999 |
| 0.887 | b = 16, r = 8 (threshold ≈ 0.71) | 1 |
| 0.9 | b = 16, r = 8 (threshold ≈ 0.71) | 1 |
| 0.912 | b = 16, r = 8 (threshold ≈ 0.71) | 1 |
| 0.925 | b = 16, r = 8 (threshold ≈ 0.71) | 1 |
| 0.938 | b = 16, r = 8 (threshold ≈ 0.71) | 1 |
| 0.95 | b = 16, r = 8 (threshold ≈ 0.71) | 1 |
| 0.963 | b = 16, r = 8 (threshold ≈ 0.71) | 1 |
| 0.975 | b = 16, r = 8 (threshold ≈ 0.71) | 1 |
| 0.988 | b = 16, r = 8 (threshold ≈ 0.71) | 1 |
| 1 | b = 16, r = 8 (threshold ≈ 0.71) | 1 |
| 0 | b = 32, r = 4 (threshold ≈ 0.42) | 0 |
| 0.013 | b = 32, r = 4 (threshold ≈ 0.42) | 0 |
| 0.025 | b = 32, r = 4 (threshold ≈ 0.42) | 0 |
| 0.038 | b = 32, r = 4 (threshold ≈ 0.42) | 0 |
| 0.05 | b = 32, r = 4 (threshold ≈ 0.42) | 0 |
| 0.063 | b = 32, r = 4 (threshold ≈ 0.42) | 0 |
| 0.075 | b = 32, r = 4 (threshold ≈ 0.42) | 0.001 |
| 0.088 | b = 32, r = 4 (threshold ≈ 0.42) | 0.002 |
| 0.1 | b = 32, r = 4 (threshold ≈ 0.42) | 0.003 |
| 0.113 | b = 32, r = 4 (threshold ≈ 0.42) | 0.005 |
| 0.125 | b = 32, r = 4 (threshold ≈ 0.42) | 0.008 |
| 0.138 | b = 32, r = 4 (threshold ≈ 0.42) | 0.011 |
| 0.15 | b = 32, r = 4 (threshold ≈ 0.42) | 0.016 |
| 0.163 | b = 32, r = 4 (threshold ≈ 0.42) | 0.022 |
| 0.175 | b = 32, r = 4 (threshold ≈ 0.42) | 0.03 |
| 0.188 | b = 32, r = 4 (threshold ≈ 0.42) | 0.039 |
| 0.2 | b = 32, r = 4 (threshold ≈ 0.42) | 0.05 |
| 0.212 | b = 32, r = 4 (threshold ≈ 0.42) | 0.063 |
| 0.225 | b = 32, r = 4 (threshold ≈ 0.42) | 0.079 |
| 0.237 | b = 32, r = 4 (threshold ≈ 0.42) | 0.097 |
| 0.25 | b = 32, r = 4 (threshold ≈ 0.42) | 0.118 |
| 0.263 | b = 32, r = 4 (threshold ≈ 0.42) | 0.141 |
| 0.275 | b = 32, r = 4 (threshold ≈ 0.42) | 0.168 |
| 0.287 | b = 32, r = 4 (threshold ≈ 0.42) | 0.197 |
| 0.3 | b = 32, r = 4 (threshold ≈ 0.42) | 0.229 |
| 0.313 | b = 32, r = 4 (threshold ≈ 0.42) | 0.264 |
| 0.325 | b = 32, r = 4 (threshold ≈ 0.42) | 0.302 |
| 0.338 | b = 32, r = 4 (threshold ≈ 0.42) | 0.342 |
| 0.35 | b = 32, r = 4 (threshold ≈ 0.42) | 0.384 |
| 0.362 | b = 32, r = 4 (threshold ≈ 0.42) | 0.427 |
| 0.375 | b = 32, r = 4 (threshold ≈ 0.42) | 0.472 |
| 0.388 | b = 32, r = 4 (threshold ≈ 0.42) | 0.518 |
| 0.4 | b = 32, r = 4 (threshold ≈ 0.42) | 0.564 |
| 0.412 | b = 32, r = 4 (threshold ≈ 0.42) | 0.609 |
| 0.425 | b = 32, r = 4 (threshold ≈ 0.42) | 0.654 |
| 0.438 | b = 32, r = 4 (threshold ≈ 0.42) | 0.697 |
| 0.45 | b = 32, r = 4 (threshold ≈ 0.42) | 0.738 |
| 0.463 | b = 32, r = 4 (threshold ≈ 0.42) | 0.777 |
| 0.475 | b = 32, r = 4 (threshold ≈ 0.42) | 0.812 |
| 0.487 | b = 32, r = 4 (threshold ≈ 0.42) | 0.844 |
| 0.5 | b = 32, r = 4 (threshold ≈ 0.42) | 0.873 |
| 0.512 | b = 32, r = 4 (threshold ≈ 0.42) | 0.898 |
| 0.525 | b = 32, r = 4 (threshold ≈ 0.42) | 0.92 |
| 0.537 | b = 32, r = 4 (threshold ≈ 0.42) | 0.939 |
| 0.55 | b = 32, r = 4 (threshold ≈ 0.42) | 0.954 |
| 0.563 | b = 32, r = 4 (threshold ≈ 0.42) | 0.966 |
| 0.575 | b = 32, r = 4 (threshold ≈ 0.42) | 0.975 |
| 0.588 | b = 32, r = 4 (threshold ≈ 0.42) | 0.983 |
| 0.6 | b = 32, r = 4 (threshold ≈ 0.42) | 0.988 |
| 0.613 | b = 32, r = 4 (threshold ≈ 0.42) | 0.992 |
| 0.625 | b = 32, r = 4 (threshold ≈ 0.42) | 0.995 |
| 0.637 | b = 32, r = 4 (threshold ≈ 0.42) | 0.997 |
| 0.65 | b = 32, r = 4 (threshold ≈ 0.42) | 0.998 |
| 0.662 | b = 32, r = 4 (threshold ≈ 0.42) | 0.999 |
| 0.675 | b = 32, r = 4 (threshold ≈ 0.42) | 0.999 |
| 0.688 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.7 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.713 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.725 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.738 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.75 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.762 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.775 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.787 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.8 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.813 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.825 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.838 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.85 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.863 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.875 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.887 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.9 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.912 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.925 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.938 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.95 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.963 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.975 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 0.988 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
| 1 | b = 32, r = 4 (threshold ≈ 0.42) | 1 |
Worked examples, recomputed
Every number the mechanisms quote for this relationship is recomputed from it, so none of them disagrees with it.
Where it appears
Each mechanism states it in its own names. The general symbol each name stands for is shown with it.
- Core Data Structures
How do we choose a data structure when the real cost is set by the hardware, not by Big-O?