info.debatty.java.graphs.build

Class PartitioningGraphBuilder<T>

java.lang.Object

info.debatty.java.graphs.build.GraphBuilder<T>

info.debatty.java.graphs.build.PartitioningGraphBuilder<T>

Type Parameters:

T - The type of nodes value

All Implemented Interfaces:

Serializable

Direct Known Subclasses:

NNCTPH

public abstract class PartitioningGraphBuilder<T>
extends GraphBuilder<T>

Abstract class for graph building algorithms that split the dataset into partitions (for example using LSH). The number of stages (n_stages) and the number of partitions (n_partitions) allow you to control the speedup and precision. If brute-force method is used inside the partitions: number of similarities to compute ≃ n² / 2 x n_stages / n_partitions where n is the size of the dataset Thus speedup with respect to pure brute force ≃ n_partitions / n_stages At the same time: - increasing n_stages will increase the precision - increasing n_partitions will decrease the precision The exact relation between precision and these 2 parameters depends on the algorithm used...

See Also:

Serialized Form

Field Summary

Fields inherited from class info.debatty.java.graphs.build.GraphBuilder

DEFAULT_K

Constructor Summary

Constructors

Constructor and Description
PartitioningGraphBuilder()

Method Summary

Modifier and Type	Method and Description
protected Graph<T>	computeGraph(List<T> nodes, int k, SimilarityInterface<T> similarity)
GraphBuilder	getInternalBuilder()
int	getNPartitions() Default = 4.
int	getOversampling()
protected abstract List<T>[]	partition(List<T> nodes)
void	setInternalBuilder(GraphBuilder internal_builder) Default = Brute force.
void	setNPartitions(int n_partitions) Set the number of partitions to build for each stage.
void	setOversampling(int oversampling) Default = 2.

Methods inherited from class info.debatty.java.graphs.build.GraphBuilder

computeGraph, getComputedSimilarities, getK, getSimilarity, readFile, setK, setSimilarity, test

Methods inherited from class java.lang.Object

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

Constructor Detail

PartitioningGraphBuilder

public PartitioningGraphBuilder()

Method Detail

getOversampling

public final int getOversampling()

Returns:

setOversampling

public final void setOversampling(int oversampling)

Default = 2.

Parameters:

oversampling -

getNPartitions

public final int getNPartitions()

Default = 4.

Returns:

setNPartitions

public final void setNPartitions(int n_partitions)

Set the number of partitions to build for each stage. Attention: the number of strings per partition should be at least 100 to get relevant results! Default = 4

Parameters:

n_partitions -

getInternalBuilder

public final GraphBuilder getInternalBuilder()

Returns:

setInternalBuilder

public final void setInternalBuilder(GraphBuilder internal_builder)

Default = Brute force.

Parameters:

internal_builder -

computeGraph

protected final Graph<T> computeGraph(List<T> nodes,
                                      int k,
                                      SimilarityInterface<T> similarity)

Specified by:

computeGraph in class GraphBuilder<T>

partition

protected abstract List<T>[] partition(List<T> nodes)