com.bigdata.rdf.graph.impl.util

Class VertexDistribution

java.lang.Object

com.bigdata.rdf.graph.impl.util.VertexDistribution

public class VertexDistribution
extends Object

Utility class for sampling vertices from a graph.

If we build a table [f,v], where f is a frequency value and v is a vertex identifier, then we can select according to different biases depending on how we normalize the scores for f. For example, if f is ONE (1), and we normalize by sum(f), then we have a uniform selection over the vertices. On the other hand, if f is the #of out-edges for v and we normalize by sum(f), then we have a selection that is uniform based on the #of out-edges. Since we need to take multiple samples, the general approach is to build the table [f,v] using some policy, compute sum(f), and the select from the table by computing the desired random value in [0:1] and scanning the table until we find the corresponding row and report that vertex identifier.

Author:

Bryan Thompson

Constructor Summary

Constructors

Constructor and Description
VertexDistribution(Random r)

Method Summary

Methods

Modifier and Type	Method and Description
void	addInEdgeSample(org.openrdf.model.Resource v) Add a sample of a vertex having some in-edge.
void	addOutEdgeSample(org.openrdf.model.Resource v) Add a sample of a vertex having some out-edge.
org.openrdf.model.Resource[]	getAll() Return all (without duplicates) vertices from the graph
org.openrdf.model.Resource[]	getUnweightedSample(int desiredSampleSize, EdgesEnum edges) Return a sample (without duplicates) of vertices from the graph choosen at random without regard to their frequency distribution.
org.openrdf.model.Resource[]	getWeightedSample(int desiredSampleSize, EdgesEnum edges) Return a sample (without duplicates) of vertices from the graph choosen randomly according to their frequency within the underlying distribution.
int	size() Return the #of samples in the distribution from which a called specified number of samples may then drawn using a random sampling without replacement technique.
String	toString()

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Constructor Detail

VertexDistribution

public VertexDistribution(Random r)

Method Detail

addOutEdgeSample

public void addOutEdgeSample(org.openrdf.model.Resource v)

Add a sample of a vertex having some out-edge.

Parameters:

v - The vertex.

addInEdgeSample

public void addInEdgeSample(org.openrdf.model.Resource v)

Add a sample of a vertex having some in-edge.

Parameters:

v - The vertex.

size

public int size()

Return the #of samples in the distribution from which a called specified number of samples may then drawn using a random sampling without replacement technique.

See Also:

#getUnweightedSample(int), #getWeightedSample(int)

getWeightedSample

public org.openrdf.model.Resource[] getWeightedSample(int desiredSampleSize,
                                             EdgesEnum edges)

Return a sample (without duplicates) of vertices from the graph choosen randomly according to their frequency within the underlying distribution.

Parameters:

desiredSampleSize - The desired sample size.

edges - The sample is taken from vertices having the specified type(s) of edges. Vertices with zero degree for the specified type(s) of edges will not be present in the returned sampled.

Returns:

The distinct samples that were found.

getUnweightedSample

public org.openrdf.model.Resource[] getUnweightedSample(int desiredSampleSize,
                                               EdgesEnum edges)

Return a sample (without duplicates) of vertices from the graph choosen at random without regard to their frequency distribution.

Parameters:

desiredSampleSize - The desired sample size.

edges - The sample is taken from vertices having the specified type(s) of edges. Vertices with zero degree for the specified type(s) of edges will not be present in the returned sampled.

Returns:

The distinct samples that were found.

getAll

public org.openrdf.model.Resource[] getAll()

Return all (without duplicates) vertices from the graph

Returns:

The distinct samples that were found.

toString

public String toString()

Overrides:

toString in class Object