net.sourceforge.cilib.util.distancemeasure.CosineDistanceMeasure Maven / Gradle / Ivy
/** __ __
* _____ _/ /_/ /_ Computational Intelligence Library (CIlib)
* / ___/ / / / __ \ (c) CIRG @ UP
* / /__/ / / / /_/ / http://cilib.net
* \___/_/_/_/_.___/
*/
package net.sourceforge.cilib.util.distancemeasure;
import java.util.Collection;
import java.util.Iterator;
import net.sourceforge.cilib.type.types.Numeric;
import net.sourceforge.cilib.type.types.container.Vector;
/**
* The Cosine Distance Measure (or vector dot product) is not a distance measure, but
* rather a similarity metric. It is defined in:
* Learning structure and concepts in data through data clustering
* by Gregory James Hamerly, 2003
*
* More positive values indicate similarity. The vector dot product is the sum of the
* product of each attribute from two vectors being compared. Here we use a normalized
* version of the metric so that the dot product is always a value between -1 and 1. The
* vector dot product does not measure the difference in magnitude between two vectors;
* only the difference in the angle between the vectors. If the (normalized) result = 1,
* the two vectors are collinear and pointed in the same direction, if the (normalized)
* result = -1, the two vectors are collinear and pointed in opposite directions, and if
* the (normalized) result = 0, then the two vectors are orthogonal. We can convert this
* similarity measure to a "distance" by subtracting it from one, which will always give
* a value between 0 and 2. This is the approach we follow.
*
*/
public class CosineDistanceMeasure implements DistanceMeasure {
/**
* Calculate the "distance" (dot product) between (of) two vectors.
* @param x the one vector.
* @param y the other vector.
* @return the "distance" (or angle or dot product) (as a double) between the two vectors.
* @throws IllegalArgumentException when the two vectors' dimension differ.
* TODO: Can this not be replaced with x.dot(y)?
*/
public double distance(Collection x, Collection y) {
if (x.size() != y.size()) {
throw new IllegalArgumentException("Cannot calculate Cosine Distance for vectors of different dimensions");
}
Iterator xIterator = x.iterator();
Iterator yIterator = y.iterator();
double distance = 0.0, norm_x = 0.0, norm_y = 0.0;
double x_i = 0.0, y_i = 0.0;
for (int i = 0; i < x.size(); ++i) {
Numeric xElement = (Numeric) xIterator.next();
Numeric yElement = (Numeric) yIterator.next();
x_i = xElement.doubleValue();
y_i = yElement.doubleValue();
distance += x_i * y_i;
norm_x += x_i * x_i;
norm_y += y_i * y_i;
}
norm_x = Math.sqrt(norm_x);
norm_y = Math.sqrt(norm_y);
if (norm_x <= 0.0 || norm_y <= 0.0) {
throw new ArithmeticException("Division by zero");
}
// TODO: return x.dot(y) ???
//convert to distance by subtracting from 1
return 1.0 - (distance / (norm_x * norm_y));
}
}