Table 6 Fraction of times the estimated cosine falls within \(\pm 5\%\) of the true cosine of A and B, for different sizes of A and B. The similarities were estimated using a scale factor of 1/1000, which is the default in sourmash. In a large fraction of times, the estimated cosine is not within \(\pm 5\%\) of the true cosine
Num. elements in B | Num. elements in A | ||||
|---|---|---|---|---|---|
100K | 200K | 300K | 400K | 500K | |
100K | 0.09 | 0.20 | 0.25 | 0.32 | 0.38 |
200K | 0.17 | 0.29 | 0.55 | 0.47 | 0.51 |
300K | 0.28 | 0.40 | 0.54 | 0.58 | 0.57 |
400K | 0.32 | 0.45 | 0.61 | 0.71 | 0.74 |
500K | 0.45 | 0.42 | 0.73 | 0.66 | 0.83 |