Mirror/zeek
1
0
Fork 0
mirror of https://github.com/zeek/zeek.git synced 2025-10-04 07:38:19 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

Making the confidence configurable.

Browse source
This commit is contained in:
Robin Sommer 2013-08-31 10:23:00 -07:00
parent fb3ceae6d5
commit 295987c8d0
8 changed files with 49 additions and 40 deletions
Download patch file Download diff file Expand all files Collapse all files

43
src/probabilistic/CardinalityCounter.h
View file

@ -14,18 +14,24 @@ namespace probabilistic {
class CardinalityCounter {
public:
/**
* Constructor.
*
* Based on the error_margin, the number of buckets that need to be
* kept will be determined. Based on the max_size, the number of bits
* that will be used from the hash function will be determined.
*
* We need the hash function to return integers that are uniformly
* distributed from 0 to 2^L-1. And if that happens, the maximum
* cardinality that this counter can handle is approximately 2^L. By
* default, we will assume a value of 64 bits.
*/
CardinalityCounter(double error_margin);
* Constructor.
*
* Based on the error_margin, the number of buckets that need to be
* kept will be determined. Based on the max_size, the number of bits
* that will be used from the hash function will be determined.
*
* We need the hash function to return integers that are uniformly
* distributed from 0 to 2^L-1. And if that happens, the maximum
* cardinality that this counter can handle is approximately 2^L. By
* default, we will assume a value of 64 bits.
*
* Confidence in the estimate given by a cardinality counter is.
*
* In other words, if the cardinality is estimated to be 100 with 2%
* error margin and HLL_CONFis 0.95, then we are 95% sure that the
* actual cardinality is between 98 and 102.
*/
CardinalityCounter(double error_margin, double confidence = 0.95);
/**
* Constructor used for cloning.
@ -117,8 +123,10 @@ private:
*
* k is the number of standard deviations that we have to go to have
* a confidence level of conf.
*
* confidence: TODO.
*/
int OptimalB(double error);
int OptimalB(double error, double confidence);
/**
* Computes when the first one appears in the element. It looks at the
@ -128,15 +136,6 @@ private:
*/
uint8_t Rank(uint64 hash_modified);
/**
* Confidence in the estimate given by a cardinality counter is.
*
* In other words, if the cardinality is estimated to be 100 with 2%
* error margin and HLL_CONFis 0.95, then we are 95% sure that the
* actual cardinality is between 98 and 102.
*/
static const double HLL_CONF = .95;
/**
* This is the number of buckets that will be stored. The standard
* error is 1.04/sqrt(m), so the actual cardinality will be the