/**
* Creates a new instance of LaplaceDistribution using a weighted Maximum Likelihood estimate
* based on the given data
*
* @param data Weighed pairs of data (first is data, second is weight) that was generated by
* some unknown LaplaceDistribution distribution
* @return Maximum Likelihood UnivariateGaussian that generated the data
*/
@Override
public LaplaceDistribution learn(
final Collection<? extends WeightedValue<? extends Double>> data) {
double mean = 0.0;
double weightSum = 0.0;
for (WeightedValue<? extends Double> weightedValue : data) {
double weight = weightedValue.getWeight();
if (weight != 0.0) {
double value = weightedValue.getValue().doubleValue();
mean += weight * value;
weightSum += weight;
}
}
if (weightSum != 0.0) {
mean /= weightSum;
}
// Now compute the shape factor
double shape = 0.0;
for (WeightedValue<? extends Number> weightedValue : data) {
double weight = weightedValue.getWeight();
if (weight != 0.0) {
double value = weightedValue.getValue().doubleValue();
double delta = value - mean;
shape += weight * Math.abs(delta);
}
}
if (weightSum != 0.0) {
shape /= weightSum;
}
return new LaplaceDistribution(mean, shape);
}
/** Test of learn method, of class CategoryBalancedBaggingLearner. */
@Test
public void testLearn() {
CategoryBalancedBaggingLearner<Vector, Boolean> instance =
new CategoryBalancedBaggingLearner<Vector, Boolean>();
instance.setLearner(new Perceptron());
instance.setRandom(random);
instance.setMaxIterations(5);
instance.setPercentToSample(0.5);
assertNull(instance.getResult());
ArrayList<InputOutputPair<Vector, Boolean>> data =
new ArrayList<InputOutputPair<Vector, Boolean>>();
VectorFactory<?> vectorFactory = VectorFactory.getDefault();
for (int i = 0; i < 100; i++) {
data.add(
new DefaultInputOutputPair<Vector, Boolean>(
vectorFactory.createUniformRandom(14, 0.0, 1.0, random), true));
}
for (int i = 0; i < 2; i++) {
data.add(
new DefaultInputOutputPair<Vector, Boolean>(
vectorFactory.createUniformRandom(14, -1.0, 0.0, random), false));
}
WeightedVotingCategorizerEnsemble<Vector, Boolean, ?> result = instance.learn(data);
assertSame(result, instance.getResult());
assertEquals(5, result.getMembers().size());
for (WeightedValue<?> member : result.getMembers()) {
assertEquals(1.0, member.getWeight(), 0.0);
assertNotNull(member.getValue());
assertTrue(member.getValue() instanceof LinearBinaryCategorizer);
}
}