es.uam.eps.ir.ranksys.nn.sim

Class VectorCosineSimilarity

java.lang.Object

es.uam.eps.ir.ranksys.nn.sim.VectorSimilarity

es.uam.eps.ir.ranksys.nn.sim.VectorCosineSimilarity

All Implemented Interfaces:

Similarity

public class VectorCosineSimilarity
extends VectorSimilarity

Vector cosine similarity. As in Cremonesi's paper. Can be asymmetric if alpha != 0.5. sim(v, w) = v * w / ((v * v)^alpha (w * w)^(1 - alpha))
F. Aiolli. Efficient Top-N Recommendation for Very Large Scale Binary Rated Datasets. RecSys 2013.
P. Cremonesi, Y. Koren, and R. Turrin. Performance of recommender algorithms on top-N recommendation tasks. RecSys 2010.

Author:

Saúl Vargas (saul.vargas@uam.es)

Field Summary

Fields inherited from class es.uam.eps.ir.ranksys.nn.sim.VectorSimilarity

data, dense, norm2Array, norm2Map

Constructor Summary

Constructors

Constructor and Description
VectorCosineSimilarity(FastPreferenceData<?,?> data, double alpha, boolean dense) Constructor.

Method Summary

Modifier and Type	Method and Description
protected double	sim(double product, double norm2A, double norm2B) Calculates the similarity value.

Methods inherited from class es.uam.eps.ir.ranksys.nn.sim.VectorSimilarity

similarElems, similarity

Methods inherited from class java.lang.Object

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

Methods inherited from interface es.uam.eps.ir.ranksys.nn.sim.Similarity

similarity

Constructor Detail

VectorCosineSimilarity

public VectorCosineSimilarity(FastPreferenceData<?,?> data,
                              double alpha,
                              boolean dense)

Constructor.

Parameters:

data - preference data

alpha - asymmetry of the similarity, set to 0.5 for symmetry

dense - true for array-based calculations, false to map-based

Method Detail

sim

protected double sim(double product,
                     double norm2A,
                     double norm2B)

Description copied from class: VectorSimilarity

Calculates the similarity value.

Specified by:

sim in class VectorSimilarity

Parameters:

product - value of the inner product of vectors

norm2A - square of the norm of the first vector

norm2B - square of the norm of the second vector

Returns:

similarity value