package jsat.linear.distancemetrics;
import java.util.List;
import java.util.concurrent.ExecutorService;
import jsat.linear.Vec;
/**
* This distance metric returns the same cosine distance as
* {@link CosineDistance}. This implementation assumes that all vectors being
* passed in for distance computations have already been L2 normalized. This
* means the distance computation can be done more efficiently, but the results
* will be incorrect if the inputs have not already been normalized.
* The word Normalized is postfixed to the name to avoid confusion, as many
* might assume "Normalized-CosineDistance" would mean a cosine distance with
* some form of additional normalization.
*
* @author Edward Raff
*/
public class CosineDistanceNormalized implements DistanceMetric
{
/*
* NOTE: Math.min(val, 1) is used because numerical instability can cause
* slightly larger values than 1 when the values are extremly close to
* eachother. In this case, it would cause a negative value in the sqrt of
* the cosineToDinstance calculation, resulting in a NaN. So the max is used
* to avoid this.
*/
private static final long serialVersionUID = -4041803247001806577L;
@Override
public double dist(Vec a, Vec b)
{
return CosineDistance.cosineToDistance(Math.min(a.dot(b), 1));
}
@Override
public boolean isSymmetric()
{
return true;
}
@Override
public boolean isSubadditive()
{
return true;
}
@Override
public boolean isIndiscemible()
{
return true;
}
@Override
public double metricBound()
{
return 1;
}
@Override
public String toString()
{
return "Cosine Distance (Normalized)";
}
@Override
public CosineDistanceNormalized clone()
{
return new CosineDistanceNormalized();
}
}
Ads help maintain this website.