Example usage for org.apache.mahout.math Vector getDistanceSquared

Introduction

In this page you can find the example usage for org.apache.mahout.math Vector getDistanceSquared.

Prototype

double getDistanceSquared(Vector v);

Source Link

Document

Get the square of the distance between this vector and the other vector.

Usage

From source file:com.cloudera.science.ml.core.vectors.Centers.java

License:Open Source License

/**
 * Returns the minimum squared Euclidean distance between the given
 * {@code Vector} and a point contained in this instance.
 * /*from  w ww .j  av  a2s.  c o  m*/
 * @param point The point
 * @return The minimum squared Euclidean distance from the point 
 */
public double getDistanceSquared(Vector point) {
    double min = Double.POSITIVE_INFINITY;
    for (Vector c : centers) {
        min = Math.min(min, c.getDistanceSquared(point));
    }
    return min;
}

---

From source file:com.cloudera.science.ml.kmeans.core.KMeansEvaluation.java

License:Open Source License

private void init() {
    predictionStrengths = Lists.newArrayListWithExpectedSize(testCenters.size());
    trainCosts = Lists.newArrayListWithExpectedSize(testCenters.size());
    testCosts = Lists.newArrayListWithExpectedSize(testCenters.size());
    stableClusters = Lists.newArrayListWithExpectedSize(testCenters.size());
    stablePoints = Lists.newArrayListWithExpectedSize(testCenters.size());

    for (int i = 0; i < testCenters.size(); i++) {
        Centers test = testCenters.get(i);
        Centers train = trainCenters.get(i);
        double trainCost = 0.0, testCost = 0.0;
        double[][] assignments = new double[test.size()][train.size()];
        int totalPoints = 0;
        for (Weighted<Vector> wv : testPoints) {
            double wt = wv.weight();
            totalPoints += wt;// www . j  a  v  a 2  s . c  o  m
            Vector v = wv.thing();
            int testId = test.indexOfClosest(v);
            testCost += wt * v.getDistanceSquared(test.get(testId));
            int trainId = train.indexOfClosest(wv.thing());
            trainCost += wt * v.getDistanceSquared(train.get(trainId));
            assignments[testId][trainId] += wt;
        }
        trainCosts.add(trainCost);
        testCosts.add(testCost);

        double minScore = Double.POSITIVE_INFINITY;
        double points = 0;
        double clusters = 0;
        for (double[] assignment : assignments) {
            double total = 0.0;
            double same = 0.0;
            for (double a : assignment) {
                total += a;
                same += a * (a - 1);
            }
            double score = same / (total * (total - 1));
            // Only consider clusters that contain a non-trivial number of obs
            if (total > assignment.length && score < minScore) {
                minScore = score;
            }
            if (score > 0.8) { // stability threshold
                clusters++;
                points += total;
            }
        }
        predictionStrengths.add(minScore);
        stableClusters.add(clusters / assignments.length);
        stablePoints.add(points / totalPoints);
    }
}

---