Package de.jungblut.math

Examples of de.jungblut.math.DoubleVector

101
102
103
104
105
106
107
108
109
  /**
   * @return the average transition probability of the given sequence.
   */
  public double averageTransitionProbability(int[] sequence) {
    DoubleVector distribution = getTransitionProbabilities(sequence);
    return FastMath.exp(distribution.sum()
        / Math.max(1d, distribution.getLength()));
  }
View Full Code Here
146
147
148
149
150
151
152
153
154
155
156
      }
      if (conf.lambda != 0d) {
        thetaGradients[i] = thetaGradients[i].add((thetas[i]
            .multiply(conf.lambda / m)));
        // subtract the regularized bias
        DoubleVector regBias = thetas[i]
            .slice(0, thetas[i].getRowCount(), 0, 1).multiply(conf.lambda / m)
            .getColumnVector(0);
        thetaGradients[i].setColumnVector(0, regBias);
      }
    }
View Full Code Here
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
      LimitedPriorityQueue<VectorDistanceTuple<VALUE>> queue) {
    if (current == null) {
      return;
    }
    int s = current.splitDimension;
    DoubleVector pivot = current.keyVector;
    double distancePivotToTarget = EuclidianDistance.get().measureDistance(
        pivot, target);

    HyperRectangle leftHyperRectangle = hyperRectangle;
    HyperRectangle rightHyperRectangle = new HyperRectangle(
        hyperRectangle.min.deepCopy(), hyperRectangle.max.deepCopy());
    leftHyperRectangle.max.set(s, pivot.get(s));
    rightHyperRectangle.min.set(s, pivot.get(s));
    boolean left = target.get(s) > pivot.get(s);
    KDTreeNode nearestNode;
    HyperRectangle nearestHyperRectangle;
    KDTreeNode furtherstNode;
    HyperRectangle furtherstHyperRectangle;
    if (left) {
      nearestNode = current.left;
      nearestHyperRectangle = leftHyperRectangle;
      furtherstNode = current.right;
      furtherstHyperRectangle = rightHyperRectangle;
    } else {
      nearestNode = current.right;
      nearestHyperRectangle = rightHyperRectangle;
      furtherstNode = current.left;
      furtherstHyperRectangle = leftHyperRectangle;
    }
    getNearestNeighbourInternal(nearestNode, target, nearestHyperRectangle,
        maxDistSquared, k, radius, queue);

    double distanceSquared = queue.isFull() ? queue.getMaximumPriority()
        : Double.MAX_VALUE;
    maxDistSquared = Math.min(maxDistSquared, distanceSquared);
    DoubleVector closest = furtherstHyperRectangle.closestPoint(target);
    double closestDistance = EuclidianDistance.get().measureDistance(closest,
        target);
    // check subtrees even if they aren't in your maxDist but within our radius
    if (closestDistance < maxDistSquared || closestDistance < radius) {
      if (distancePivotToTarget < distanceSquared) {
View Full Code Here
77
78
79
80
81
82
83
84
85
86
87
88
89
90
  /**
   * Merges two centers when they are within the given distance of each other.
   */
  private static void merge(List<DoubleVector> centers, double mergeWindow) {
    for (int i = 0; i < centers.size(); i++) {
      DoubleVector referenceVector = centers.get(i);
      // find centers to merge if they are within our merge window
      for (int j = i + 1; j < centers.size(); j++) {
        DoubleVector center = centers.get(j);
        double dist = EuclidianDistance.get().measureDistance(referenceVector,
            center);
        if (dist < mergeWindow) {
          centers.remove(j);
          centers.set(i, referenceVector.add(center).divide(2d));
View Full Code Here
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
   */
  private static int meanShift(KDTree<Integer> kdTree,
      List<DoubleVector> centers, double h) {
    int remainingConvergence = 0;
    for (int i = 0; i < centers.size(); i++) {
      DoubleVector v = centers.get(i);
      List<VectorDistanceTuple<Integer>> neighbours = kdTree
          .getNearestNeighbours(v, h);
      double weightSum = 0d;
      DoubleVector numerator = new DenseDoubleVector(v.getLength());
      for (VectorDistanceTuple<Integer> neighbour : neighbours) {
        if (neighbour.getDistance() < h) {
          double normDistance = neighbour.getDistance() / h;
          weightSum -= gaussianGradient(normDistance);
          numerator = numerator.add(neighbour.getVector().multiply(weightSum));
        }
      }
      if (weightSum > 0d) {
        DoubleVector shift = v.divide(numerator);
        DoubleVector newCenter = v.add(shift);
        if (v.subtract(newCenter).abs().sum() > 1e-5) {
          remainingConvergence++;
          // apply the shift
          centers.set(i, newCenter);
        }
View Full Code Here
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
    BitSet assignedIndices = new BitSet(kdTree.size());
    // we are doing one pass over the dataset to determine the first centers
    // that are within h range
    for (int i = 0; i < points.size(); i++) {
      if (!assignedIndices.get(i)) {
        DoubleVector v = points.get(i);
        List<VectorDistanceTuple<Integer>> neighbours = kdTree
            .getNearestNeighbours(v, h);
        DoubleVector center = new DenseDoubleVector(v.getLength());
        int added = 0;
        for (VectorDistanceTuple<Integer> neighbour : neighbours) {
          if (!assignedIndices.get(neighbour.getValue())
              && neighbour.getDistance() < h) {
            center = center.add(neighbour.getVector());
            assignedIndices.set(neighbour.getValue());
            added++;
          }
        }
        // so if our sum is positive, we can divide and add the center
        if (added > 1) {
          DoubleVector newCenter = center.divide(added);
          centers.add(newCenter);
          if (verbose && centers.size() % 1000 == 0) {
            LOG.info("#Centers found: " + centers.size());
          }
        }
View Full Code Here
5
6
7
8
9
10
11
12
public abstract class AbstractPredictor implements Predictor {

  @Override
  public int predictedClass(DoubleVector features, double threshold) {
    DoubleVector predict = predict(features);
    return extractPredictedClass(predict, threshold);
  }
View Full Code Here
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
*/
public abstract class AbstractActivationFunction implements ActivationFunction {

  @Override
  public DoubleVector apply(DoubleVector vector) {
    DoubleVector newInstance = newInstance(vector);
    if (vector.isSparse()) {
      Iterator<DoubleVectorElement> iterateNonZero = vector.iterateNonZero();
      while (iterateNonZero.hasNext()) {
        DoubleVectorElement next = iterateNonZero.next();
        newInstance.set(next.getIndex(), apply(next.getValue()));
      }
    } else {
      for (int i = 0; i < vector.getDimension(); i++) {
        newInstance.set(i, apply(vector.get(i)));
      }
    }
    return newInstance;
  }
View Full Code Here
11
12
13
14
15
16
17
18
    return extractPredictedClass(predict, threshold);
  }

  @Override
  public int predictedClass(DoubleVector features) {
    DoubleVector predict = predict(features);
    return extractPredictedClass(predict);
  }
View Full Code Here
17
18
19
20
21
22
23
24
    return extractPredictedClass(predict);
  }

  @Override
  public DoubleVector predictProbability(DoubleVector features) {
    DoubleVector predict = predict(features);
    return predict.divide(predict.sum());
  }
View Full Code Here
TOP

Related Classes of de.jungblut.math.DoubleVector

  • de.jungblut.classification.AbstractPredictor
  • de.jungblut.classification.bayes.MultinomialNaiveBayes
  • de.jungblut.classification.bayes.MultinomialNaiveBayesTest
  • de.jungblut.classification.eval.Evaluator
  • de.jungblut.classification.knn.AbstractKNearestNeighbours
  • de.jungblut.classification.knn.KNearestNeighboursTest
  • de.jungblut.classification.knn.SparseKNearestNeighboursTest
  • de.jungblut.classification.meta.Voter
  • de.jungblut.classification.nn.MultilayerPerceptron
  • de.jungblut.classification.nn.MultilayerPerceptronCostFunction
    • Copyright © 2018 www.massapicom. All rights reserved.
    All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.