HashBucket Class Reference

#include <HashBucket.h>

Public Member Functions
	HashBucket ()
	HashBucket (VectorXd p)
	HashBucket (VectorXd p, double wi)
	HashBucket (VectorXd p, double wi, int idx, int nscales)
void	update (VectorXd p, std::mt19937_64 &rng)
void	update (VectorXd p, double wi, std::mt19937_64 &rng)
void	update (VectorXd p, double wi, int idx, std::mt19937_64 &rng)

Public Attributes
int	SCALES = 1
vector< int >	count
vector< VectorXd >	sample
vector< double >	wSum
std::uniform_real_distribution	unif

Detailed Description

Hash bucket for LSH tables

Constructor & Destructor Documentation

HashBucket() [1/4]

HashBucket::HashBucket ( )

inline

Default constructor.

HashBucket() [2/4]

HashBucket::HashBucket ( VectorXd  p )

inline

Build a new hash bucket for data point p.

Parameters

p	a data point that belongs to the bucket

HashBucket() [3/4]

HashBucket::HashBucket ( VectorXd  p, double  wi  )

inline

Build a new hash bucket for data point p and weight wi.

Parameters

p	a data point that belongs to the bucket
wi	weight of the data point

HashBucket() [4/4]

HashBucket::HashBucket ( VectorXd  p, double  wi, int  idx, int  nscales  )

inline

Build a new hash bucket for data point p, weight wi, scale idx out of a total of nscales scales

Parameters

p	data
wi	weight
idx	index of scale that p falls into in the bucket
nscales	total number of scales

Member Function Documentation

update() [1/3]

void HashBucket::update ( VectorXd  p, std::mt19937_64 &  rng  )

inline

Insert point into bucket: update count and replace bucket sample with reservoir sampling.

Parameters

p	data point
rng	random number generator

update() [2/3]

void HashBucket::update ( VectorXd  p, double  wi, std::mt19937_64 &  rng  )

inline

Insert point p with weight wi into bucket: update count and replace bucket sample with weighted reservoir sampling (A-Chao).

Parameters

p	data point
wi	weight of data point
rng	random number generator

update() [3/3]

void HashBucket::update ( VectorXd  p, double  wi, int  idx, std::mt19937_64 &  rng  )

inline

Insert point p with weight wi, belonging to scale idx into bucket: update count and replace bucket sample with weighted reservoir sampling (A-Chao). Note that samples for each scale is stored and updated separately.

Parameters

p	data point
wi	weight of data point
idx	scale idx of data point
rng	random number generator

Member Data Documentation

count

vector<int> HashBucket::count

Count of data points that falls into the bucket

sample

vector<VectorXd> HashBucket::sample

Samples of data points that falls into the bucket

SCALES

int HashBucket::SCALES = 1

Store SCALES data samples per bucket, one from each weight scale. Comparing to storing 1 sample per bucket (default), storing more samples slightly longer but is more accurate. We do not recommend setting this to more than 3.

wSum

vector<double> HashBucket::wSum

Sum of weights of all data points that falls into the bucket.

---

The documentation for this class was generated from the following file:

• HashBucket.h