You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I believe the accuracy variable used in the determining the rows parameter should be actually 1 - accuracy. This is based on the this documentation.
More specifically this paragraph which states that for a probability at least 1−δ, the CountMinSketch algorithm will require ln 1/δ rows.
There's also the fact that, at the moment, ln 1/δ with an δ of 0.999 will result in 1 rows for the algorithm, but that's counter intuitive because for a better accuracy, we will need more rows, not less.
The text was updated successfully, but these errors were encountered:
Hi,
I noticed that the documentation on CountMinSketch is wrong. It states that the
accuracy
parameter to be used should be0.999
(which is also the default value at https://github.com/Callidon/bloom-filters/blob/eb43a770544e1796e8e04919a30bd7cb2bf09e0f/src/sketch/count-min-sketch.ts#L77C54-L77C59)I believe the
accuracy
variable used in the determining therows
parameter should be actually1 - accuracy
. This is based on the this documentation.More specifically this paragraph which states that for a
probability at least 1−δ
, the CountMinSketch algorithm will requireln 1/δ
rows.There's also the fact that, at the moment,
ln 1/δ
with anδ
of0.999
will result in1
rows for the algorithm, but that's counter intuitive because for a better accuracy, we will need more rows, not less.The text was updated successfully, but these errors were encountered: