ai.libs.jaicore.ml.core.dataset.attribute.transformer.multivalue

Class MinHashingTransformer

java.lang.Object

ai.libs.jaicore.ml.core.dataset.attribute.transformer.multivalue.MinHashingTransformer

All Implemented Interfaces:

ISingleAttributeTransformer

public class MinHashingTransformer
extends java.lang.Object
implements ISingleAttributeTransformer

Converts the sets of multi-value features to short signatures. At first the feature value is transformed into a binaryzation, i.e. a 0/1 Vector, and the MinHashing applied on this vectors afterwards. If two multi-value feature sets are very similar with respect to the Jaccard-Similarity, then the two signatures will be similar as well with a high probability depending on the desired length of the signatures.
For a signature of length n, the same amount of permutations will be created and the n-th element of the signature is determined by the index where the n-th permutation finds the finds the first 1 in the 0/1 Vector.

Constructor Summary

Constructors

Constructor and Description
MinHashingTransformer(int[][] permutations) Constructor where the user gives predefined permutations.
MinHashingTransformer(int domainSize, int signatureLength, long seed) Constructor where suitable permutations are created randomly.

Method Summary

Modifier and Type	Method and Description
double[]	transformAttribute(IAttributeValue<?> attributeToTransform)

Methods inherited from class java.lang.Object

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

Constructor Detail

MinHashingTransformer

public MinHashingTransformer(int[][] permutations)

Constructor where the user gives predefined permutations.

Parameters:

permutations - Predefined permutations. The amount of permutations defines the length of the signature the MinHashing creates and each permutation has to have the length of the domain size.

MinHashingTransformer

public MinHashingTransformer(int domainSize,
                             int signatureLength,
                             long seed)

Constructor where suitable permutations are created randomly.

Parameters:

domainSize -

signatureLength -

Method Detail

transformAttribute

public double[] transformAttribute(IAttributeValue<?> attributeToTransform)

Specified by:

transformAttribute in interface ISingleAttributeTransformer