@Override
public double estimate(Vector example, Object label) {
Objects.requireNonNull(example, "Require an example.");
int trueClassIndex = classifier.getClasses().loc().indexOf(label);
Check.argument(trueClassIndex >= 0, "illegal label %s", label);
return errorFunction.apply(classifier.estimate(example), trueClassIndex);
}
@Override
public DoubleArray estimate(DataFrame x, Vector y) {
Objects.requireNonNull(x, "Input data required.");
Objects.requireNonNull(y, "Input target required.");
Check.argument(x.rows() == y.size(), "The size of input data and input target don't match.");
return errorFunction.apply(classifier.estimate(x), y, classifier.getClasses());
}
@Override
public InductiveConformalClassifier fit(DataFrame x, Vector y) {
Objects.requireNonNull(x, "Input data is required.");
Objects.requireNonNull(y, "Input target is required.");
Check.argument(x.rows() == y.size(), "The size of input data and input target don't match.");
return new InductiveConformalClassifier(learner.fit(x, y), Vectors.unique(y));
}
@Override
public Nonconformity fit(DataFrame x, Vector y) {
Objects.requireNonNull(x, "Input data is required.");
Objects.requireNonNull(y, "Input target is required.");
Check.argument(x.rows() == y.size(), "The size of input data and input target don't match");
Classifier probabilityEstimator = classifier.fit(x, y);
Check.state(
probabilityEstimator != null
&& probabilityEstimator
.getCharacteristics()
.contains(ClassifierCharacteristic.ESTIMATOR),
"The produced classifier can't estimate probabilities");
return new ProbabilityEstimateNonconformity(probabilityEstimator, errorFunction);
}