org.aksw.limes.core.measures.mapper.space.blocking

Class EuclideanBlockingModule

java.lang.Object

org.aksw.limes.core.measures.mapper.space.blocking.EuclideanBlockingModule

All Implemented Interfaces:

IBlockingModule

public class EuclideanBlockingModule
extends Object
implements IBlockingModule

Author:

Axel-C. Ngonga Ngomo (ngonga@informatik.uni-leipzig.de)

Constructor Summary

Constructors

Constructor and Description
EuclideanBlockingModule(String props, String measureName, double threshold) Initializes the generator.

Method Summary

Modifier and Type	Method and Description
static ArrayList<ArrayList<Double>>	addIdsToList(ArrayList<ArrayList<Double>> keys, TreeSet<String> propValues)
static ArrayList<Double>	copyList(ArrayList<Double> list)
ArrayList<ArrayList<Integer>>	getAllBlockIds(Instance a)
ArrayList<ArrayList<Integer>>	getAllSourceIds(Instance a, String sourceProps)
ArrayList<Integer>	getBlockId(Instance a) Computes the block ID for a given instance a.
ArrayList<ArrayList<Integer>>	getBlocksToCompare(ArrayList<Integer> blockId) Computes the ids of all the blocks surrounding a given block for comparison Will be extremely useful for parallelizing as we can use blocking on T and S as then put use locality

Methods inherited from class java.lang.Object

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

Constructor Detail

EuclideanBlockingModule

public EuclideanBlockingModule(String props,
                               String measureName,
                               double threshold)

Initializes the generator. The basic idea here is the following: First, pick a random instance origin. That is the center upon which the block ids will be computed. Each measure can return the threshold for blocking that is equivalent to the similarity threshold given in by the user. For euclidean metrics, this value is the same. Yet, for metrics that squeeze space, this might not be the case. It is important to notice that the generation assumes that the size of props.split("|") is the same as dimensions.

Parameters:

props - List of properties that make up each dimension

measureName - Name of the measure to be used to compute the similarity of instances

threshold - General similarity threshold for the metric. This threshold is transformed into a distance threshold, as sim = a, d = (1 - a)/a. The space tiling is carried out according to distances, not similarities. Still, we can ensure that all points within the similarity range are found.

Method Detail

copyList

public static ArrayList<Double> copyList(ArrayList<Double> list)

addIdsToList

public static ArrayList<ArrayList<Double>> addIdsToList(ArrayList<ArrayList<Double>> keys,
                                                        TreeSet<String> propValues)

getBlockId

public ArrayList<Integer> getBlockId(Instance a)

Computes the block ID for a given instance a. The idea behind the blocking is to tile the target space into blocks of dimension thresdhold^dimensions. Each instance s from the source space is then compared with the blocks lying directly around s's block and the block where s is.

Specified by:

getBlockId in interface IBlockingModule

Parameters:

a - The instance whose blockId is to be computed

Returns:

The ID for the block of a

getBlocksToCompare

public ArrayList<ArrayList<Integer>> getBlocksToCompare(ArrayList<Integer> blockId)

Computes the ids of all the blocks surrounding a given block for comparison Will be extremely useful for parallelizing as we can use blocking on T and S as then put use locality

Specified by:

getBlocksToCompare in interface IBlockingModule

Parameters:

blockId - index

Returns:

the ids of all the blocks surrounding a given block for comparison

getAllBlockIds

public ArrayList<ArrayList<Integer>> getAllBlockIds(Instance a)

Specified by:

getAllBlockIds in interface IBlockingModule

getAllSourceIds

public ArrayList<ArrayList<Integer>> getAllSourceIds(Instance a,
                                                     String sourceProps)

Specified by:

getAllSourceIds in interface IBlockingModule