index
int64 | repo_id
string | file_path
string | content
string |
|---|---|---|---|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/AreaUnderROCCurve.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
public class AreaUnderROCCurve extends AAreaUnderCurvePerformanceMeasure {
public AreaUnderROCCurve(final int positiveClass) {
super(positiveClass);
}
@Override
public double getXValue(final int tp, final int fp, final int tn, final int fn) {
// false positive rate
return (double) fp / (fp + tn);
}
@Override
public double getYValue(final int tp, final int fp, final int tn, final int fn) {
// true positive rate
return (double) tp / (tp + fn);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/AveragedInstanceLoss.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
import java.util.List;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
import org.api4.java.ai.ml.core.evaluation.supervised.loss.IDeterministicInstancePredictionPerformanceMeasure;
public class AveragedInstanceLoss extends ASingleLabelClassificationPerformanceMeasure {
private IDeterministicInstancePredictionPerformanceMeasure<ISingleLabelClassification, Integer> instanceMeasure;
public AveragedInstanceLoss(final IDeterministicInstancePredictionPerformanceMeasure<ISingleLabelClassification, Integer> instanceMeasure) {
this.instanceMeasure = instanceMeasure;
}
@Override
public double loss(final List<? extends Integer> expected, final List<? extends ISingleLabelClassification> predicted) {
return IntStream.range(0, expected.size()).mapToDouble(x -> this.instanceMeasure.loss(expected.get(x), predicted.get(x))).average().getAsDouble();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/EAggregatedClassifierMetric.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
import java.util.ArrayList;
import java.util.List;
import java.util.stream.Collectors;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
import org.api4.java.ai.ml.core.evaluation.IPredictionAndGroundTruthTable;
import org.api4.java.ai.ml.core.evaluation.execution.IAggregatedPredictionPerformanceMeasure;
import org.api4.java.ai.ml.core.evaluation.supervised.loss.IDeterministicPredictionPerformanceMeasure;
import org.api4.java.common.aggregate.IAggregateFunction;
import ai.libs.jaicore.basic.aggregate.reals.Mean;
public enum EAggregatedClassifierMetric implements IAggregatedPredictionPerformanceMeasure<Integer, ISingleLabelClassification> {
MEAN_ERRORRATE(EClassificationPerformanceMeasure.ERRORRATE, new Mean());
private final IDeterministicPredictionPerformanceMeasure<Integer, ISingleLabelClassification> lossFunction;
private final IAggregateFunction<Double> aggregation;
private EAggregatedClassifierMetric(final IDeterministicPredictionPerformanceMeasure<Integer, ISingleLabelClassification> lossFunction, final IAggregateFunction<Double> aggregation) {
this.lossFunction = lossFunction;
this.aggregation = aggregation;
}
@Override
public double loss(final List<List<? extends Integer>> expected, final List<List<? extends ISingleLabelClassification>> predicted) {
int n = expected.size();
List<Double> losses = new ArrayList<>();
for (int i = 0; i < n; i++) {
losses.add(this.lossFunction.loss(expected.get(i), predicted.get(i)));
}
return this.aggregation.aggregate(losses);
}
@Override
public double loss(final List<IPredictionAndGroundTruthTable<? extends Integer, ? extends ISingleLabelClassification>> pairTables) {
return this.aggregation.aggregate(pairTables.stream().map(this.lossFunction::loss).collect(Collectors.toList()));
}
@Override
public double score(final List<List<? extends Integer>> expected, final List<List<? extends ISingleLabelClassification>> predicted) {
return 1 - this.loss(expected, predicted);
}
@Override
public double score(final List<IPredictionAndGroundTruthTable<? extends Integer, ? extends ISingleLabelClassification>> pairTables) {
return 1 - this.loss(pairTables);
}
@Override
public IDeterministicPredictionPerformanceMeasure<Integer, ISingleLabelClassification> getBaseMeasure() {
return this.lossFunction;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/EClassificationPerformanceMeasure.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
import java.util.List;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
import org.api4.java.ai.ml.core.evaluation.IPredictionAndGroundTruthTable;
import org.api4.java.ai.ml.core.evaluation.supervised.loss.IDeterministicPredictionPerformanceMeasure;
public enum EClassificationPerformanceMeasure implements IDeterministicPredictionPerformanceMeasure<Integer, ISingleLabelClassification> {
// AREA_ABOVE_ROC, AREA_UNDER_ROC, AVG_COST, CORRECT, CORRELATION_COEFFICIENT, ERROR_RATE, FALSE_NEGATIVE_RATE, FALSE_POSITIVE_RATE, F_MEASURE, INCORRECT, KAPPA, KB_INFORMATION, KB_MEA_INFORMATION, KB_RELATIVE_INFORMATION,
// MEAN_ABSOLUTE_ERROR, PCT_CORRECT, PCT_INCORRECT, PRECISION, RELATIVE_ABSOLUTE_ERROR, ROOT_MEAN_SQUARED_ERROR, ROOT_RELATIVE_SQUARED_ERROR, WEIGHTED_AREA_UNDER_ROC, WEIGHTED_FALSE_NEGATIVE_RATE, WEIGHTED_FALSE_POSITIVE_RATE,
// WEIGHTED_F_MEASURE, WEIGHTED_PRECISION, WEIGHTED_RECALL, WEIGHTED_TRUE_NEGATIVE_RATE, WEIGHTED_TRUE_POSITIVE_RATE
ERRORRATE(new ErrorRate()), TRUE_NEGATIVES_WITH_1_POSITIVE(new TrueNegatives(1)), TRUE_POSITIVES_WITH_1_POSITIVE(new TruePositives(1)), FALSE_NEGATIVES_WITH_1_POSITIVE(new FalseNegatives(1)),
FALSE_POSITIVES_WITH_1_POSITIVE(new FalsePositives(1)), PRECISION_WITH_1_POSITIVE(new Precision(1)), RECALL_WITH_1_POSITIVE(new Recall(1)), F1_WITH_1_POSITIVE(new F1Measure(1));
private final IDeterministicPredictionPerformanceMeasure<Integer, ISingleLabelClassification> measure;
private EClassificationPerformanceMeasure(final IDeterministicPredictionPerformanceMeasure<Integer, ISingleLabelClassification> measure) {
this.measure = measure;
}
@Override
public double loss(final List<? extends Integer> expected, final List<? extends ISingleLabelClassification> predicted) {
return this.measure.loss(expected, predicted);
}
@Override
public double loss(final IPredictionAndGroundTruthTable<? extends Integer, ? extends ISingleLabelClassification> pairTable) {
return this.measure.loss(pairTable);
}
@Override
public double score(final List<? extends Integer> expected, final List<? extends ISingleLabelClassification> predicted) {
return this.measure.score(expected, predicted);
}
@Override
public double score(final IPredictionAndGroundTruthTable<? extends Integer, ? extends ISingleLabelClassification> pairTable) {
return this.measure.score(pairTable);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/ErrorRate.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
import java.util.List;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
public class ErrorRate extends ASingleLabelClassificationPerformanceMeasure {
public ErrorRate() {
/* empty constructor to avoid direct instantiation. Use the enum instead. */
}
@Override
public double loss(final List<? extends Integer> expected, final List<? extends ISingleLabelClassification> predicted) {
this.checkConsistency(expected, predicted);
double sumOfZOLoss = 0.0;
for (int i = 0; i < expected.size(); i++) {
sumOfZOLoss += expected.get(i).equals(predicted.get(i).getPrediction()) ? 0.0 : 1.0;
}
return sumOfZOLoss / expected.size();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/F1Measure.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
public class F1Measure extends FMeasure {
private static final int BETA = 1;
public F1Measure(final int positiveClass) {
super(BETA, positiveClass);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/FMeasure.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
import java.util.List;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
import ai.libs.jaicore.basic.metric.ConfusionMetrics;
public class FMeasure extends ASingleLabelClassificationPerformanceMeasure {
private final double beta;
private final TruePositives tp;
private final FalsePositives fp;
private final FalseNegatives fn;
public FMeasure(final double beta, final int positiveClass) {
this.beta = beta;
this.tp = new TruePositives(positiveClass);
this.fp = new FalsePositives(positiveClass);
this.fn = new FalseNegatives(positiveClass);
}
@Override
public double score(final List<? extends Integer> expected, final List<? extends ISingleLabelClassification> predicted) {
if (expected.size() != predicted.size()) {
throw new IllegalArgumentException("Expected and actual must be of the same length.");
}
return ConfusionMetrics.getFMeasure(this.beta, (int) this.tp.score(expected, predicted), (int) this.fp.score(expected, predicted), (int) this.fn.score(expected, predicted));
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/FalseNegatives.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
import java.util.List;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
public class FalseNegatives extends ASingleLabelClassificationPerformanceMeasure {
private final int positiveClass;
public FalseNegatives(final int positiveClass) {
this.positiveClass = positiveClass;
}
@Override
public double score(final List<? extends Integer> expected, final List<? extends ISingleLabelClassification> predicted) {
return IntStream.range(0, expected.size()).filter(i -> expected.get(i).equals(this.positiveClass) && !expected.get(i).equals(predicted.get(i).getPrediction())).count();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/FalsePositives.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
import java.util.List;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
public class FalsePositives extends ASingleLabelClassificationPerformanceMeasure {
private final int positiveClass;
public FalsePositives(final int positiveClass) {
this.positiveClass = positiveClass;
}
@Override
public double score(final List<? extends Integer> expected, final List<? extends ISingleLabelClassification> predicted) {
return IntStream.range(0, expected.size()).filter(i -> !expected.get(i).equals(this.positiveClass) && !expected.get(i).equals(predicted.get(i).getPrediction())).count();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/IPredictedClassPerformanceMeasure.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
import java.util.List;
public interface IPredictedClassPerformanceMeasure {
public double loss(List<? extends Integer> expected, List<? extends Integer> predicted);
public double score(List<? extends Integer> expected, List<? extends Integer> predicted);
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/Precision.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
import java.util.List;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
import ai.libs.jaicore.basic.metric.ConfusionMetrics;
public class Precision extends ASingleLabelClassificationPerformanceMeasure {
private final TruePositives tp;
private final FalsePositives fp;
public Precision(final int positiveClass) {
this.tp = new TruePositives(positiveClass);
this.fp = new FalsePositives(positiveClass);
}
@Override
public double score(final List<? extends Integer> expected, final List<? extends ISingleLabelClassification> predicted) {
return ConfusionMetrics.getPrecision((int) this.tp.score(expected, predicted), (int) this.fp.score(expected, predicted));
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/Recall.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
import java.util.List;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
import ai.libs.jaicore.basic.metric.ConfusionMetrics;
public class Recall extends ASingleLabelClassificationPerformanceMeasure {
private final TruePositives tp;
private final FalseNegatives fn;
public Recall(final int positiveClass) {
this.tp = new TruePositives(positiveClass);
this.fn = new FalseNegatives(positiveClass);
}
@Override
public double score(final List<? extends Integer> expected, final List<? extends ISingleLabelClassification> predicted) {
return ConfusionMetrics.getRecall((int) this.tp.score(expected, predicted), (int) this.fn.score(expected, predicted));
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/TrueNegatives.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
import java.util.List;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
public class TrueNegatives extends ASingleLabelClassificationPerformanceMeasure {
private final int positiveClass;
public TrueNegatives(final int positiveClass) {
this.positiveClass = positiveClass;
}
@Override
public double score(final List<? extends Integer> expected, final List<? extends ISingleLabelClassification> predicted) {
return IntStream.range(0, expected.size()).filter(i -> !expected.get(i).equals(this.positiveClass) && expected.get(i).equals(predicted.get(i).getPrediction())).count();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/TruePositives.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
import java.util.List;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
public class TruePositives extends ASingleLabelClassificationPerformanceMeasure {
private final int positiveClass;
public TruePositives(final int positiveClass) {
this.positiveClass = positiveClass;
}
@Override
public double score(final List<? extends Integer> expected, final List<? extends ISingleLabelClassification> predicted) {
return IntStream.range(0, expected.size()).filter(i -> expected.get(i).equals(this.positiveClass) && expected.get(i).equals(predicted.get(i).getPrediction())).count();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/dataset/WeightedAUROC.java
|
package ai.libs.jaicore.ml.classification.loss.dataset;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Map.Entry;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
import ai.libs.jaicore.basic.Maps;
/**
* Computes the AUROC weighted by class sizes, that is, it first computes the size of each class,
* then computes AUROC in a one-vs-rest fashion and balances the final score proportional to the
* size of each class.
*
* @author mwever
*/
public class WeightedAUROC extends ASingleLabelClassificationPerformanceMeasure {
@Override
public double score(final List<? extends Integer> expected, final List<? extends ISingleLabelClassification> predicted) {
Map<Integer, Integer> classCountMap = new HashMap<>();
expected.stream().forEach(x -> Maps.increaseCounterInMap(classCountMap, x));
double sum = 0;
for (Entry<Integer, Integer> posClassEntry : classCountMap.entrySet()) {
sum += new AreaUnderROCCurve(posClassEntry.getKey()).score(expected, predicted) * posClassEntry.getValue();
}
return sum / classCountMap.values().stream().mapToInt(x -> x).sum();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/instance/AInstanceMeasure.java
|
package ai.libs.jaicore.ml.classification.loss.instance;
import org.api4.java.ai.ml.core.evaluation.supervised.loss.IDeterministicInstancePredictionPerformanceMeasure;
/**
* Abstract class for instance-based measures.
*
* @author mwever
*
* @param <E> The type of the expected value.
* @param <A> The type of the actual/predicted.
*/
public class AInstanceMeasure<E, A> implements IDeterministicInstancePredictionPerformanceMeasure<A, E> {
@Override
public double loss(final E expected, final A predicted) {
return 1 - this.score(expected, predicted);
}
@Override
public double score(final E expected, final A predicted) {
return 1 - this.loss(expected, predicted);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/instance/CrossEntropyLoss.java
|
package ai.libs.jaicore.ml.classification.loss.instance;
import java.util.Map;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
public class CrossEntropyLoss extends AInstanceMeasure<double[], ISingleLabelClassification> {
public static final double DEF_EPSILON = 1E-15;
private final double epsilon;
public CrossEntropyLoss() {
this(DEF_EPSILON);
}
public CrossEntropyLoss(final double epsilon) {
this.epsilon = epsilon;
}
@Override
public double loss(final double[] expected, final ISingleLabelClassification predicted) {
Map<Integer, Double> distributionMap = predicted.getClassDistribution();
double[] predictedArr = new double[distributionMap.size()];
IntStream.range(0, distributionMap.size()).forEach(x -> predictedArr[x] = distributionMap.get(x));
return -IntStream.range(0, expected.length).mapToDouble(i -> expected[i] * Math.log(this.minMax(predictedArr[i]))).sum();
}
private double minMax(final double value) {
return Math.min(1 - this.epsilon, Math.max(value, this.epsilon));
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/instance/JaccardScore.java
|
package ai.libs.jaicore.ml.classification.loss.instance;
import java.util.Collection;
import ai.libs.jaicore.basic.sets.SetUtil;
public class JaccardScore extends AInstanceMeasure<Collection<Integer>, Collection<Integer>> {
@Override
public double score(final Collection<Integer> expected, final Collection<Integer> predicted) {
return ((double) SetUtil.intersection(expected, predicted).size()) / SetUtil.union(expected, predicted).size();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/instance/LogLoss.java
|
package ai.libs.jaicore.ml.classification.loss.instance;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
public class LogLoss extends AInstanceMeasure<Integer, ISingleLabelClassification> {
private final CrossEntropyLoss cel;
public LogLoss() {
this.cel = new CrossEntropyLoss();
}
public LogLoss(final double epsilon) {
this.cel = new CrossEntropyLoss(epsilon);
}
@Override
public double loss(final Integer expected, final ISingleLabelClassification predicted) {
double[] expectedArr = new double[predicted.getClassDistribution().size()];
expectedArr[expected] = 1.0;
return this.cel.loss(expectedArr, predicted);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/loss/instance/ZeroOneLoss.java
|
package ai.libs.jaicore.ml.classification.loss.instance;
import org.api4.java.ai.ml.core.evaluation.supervised.loss.IDeterministicInstancePredictionPerformanceMeasure;
/**
*
* @author mwever
*/
public class ZeroOneLoss implements IDeterministicInstancePredictionPerformanceMeasure<Object, Object> {
@Override
public double loss(final Object expected, final Object predicted) {
return expected.equals(predicted) ? 0.0 : 1.0;
}
@Override
public double score(final Object expected, final Object predicted) {
return 1 - this.loss(expected, predicted);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/MultiLabelClassification.java
|
package ai.libs.jaicore.ml.classification.multilabel;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import ai.libs.jaicore.ml.core.evaluation.Prediction;
public class MultiLabelClassification extends Prediction implements IMultiLabelClassification {
private static final double DEFAULT_THRESHOLD = 0.5;
private double[] threshold;
public MultiLabelClassification(final double[] predicted) {
this(predicted, DEFAULT_THRESHOLD);
}
public MultiLabelClassification(final double[] predicted, final double threshold) {
this(predicted, IntStream.range(0, predicted.length).mapToDouble(x -> threshold).toArray());
}
public MultiLabelClassification(final double[] predicted, final double[] threshold) {
super(predicted);
this.threshold = threshold;
}
@Override
public double[] getPrediction() {
return (double[]) super.getPrediction();
}
public int[] getThresholdedPrediction() {
return IntStream.range(0, this.getPrediction().length).map(x -> this.getPrediction()[x] >= this.threshold[x] ? 1 : 0).toArray();
}
@Override
public int[] getPrediction(final double threshold) {
return IntStream.range(0, this.getPrediction().length).map(x -> this.getPrediction()[x] >= threshold ? 1 : 0).toArray();
}
@Override
public int[] getPrediction(final double[] threshold) {
return IntStream.range(0, this.getPrediction().length).map(x -> this.getPrediction()[x] >= threshold[x] ? 1 : 0).toArray();
}
@Override
public int[] getRelevantLabels(final double threshold) {
return IntStream.range(0, this.getPrediction().length).filter(x -> this.getPrediction()[x] >= threshold).toArray();
}
@Override
public int[] getIrrelevantLabels(final double threshold) {
return IntStream.range(0, this.getPrediction().length).filter(x -> this.getPrediction()[x] < threshold).toArray();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/MultiLabelClassificationPredictionBatch.java
|
package ai.libs.jaicore.ml.classification.multilabel;
import java.util.Arrays;
import java.util.List;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassificationPredictionBatch;
public class MultiLabelClassificationPredictionBatch implements IMultiLabelClassificationPredictionBatch {
private List<? extends IMultiLabelClassification> batch;
public MultiLabelClassificationPredictionBatch(final List<? extends IMultiLabelClassification> batch) {
this.batch = batch;
}
public MultiLabelClassificationPredictionBatch(final IMultiLabelClassification[] batch) {
this(Arrays.asList(batch));
}
@Override
public IMultiLabelClassification get(final int index) {
return this.batch.get(index);
}
@Override
public int getNumPredictions() {
return this.batch.size();
}
@Override
public List<? extends IMultiLabelClassification> getPredictions() {
return this.batch;
}
@Override
public double[][] getPredictionMatrix() {
double[][] predictionMatrix = new double[this.batch.size()][];
IntStream.range(0, this.batch.size()).forEach(x -> predictionMatrix[x] = this.batch.get(x).getPrediction());
return predictionMatrix;
}
@Override
public int[][] getThresholdedPredictionMatrix(final double threshold) {
int[][] predictionMatrix = new int[this.batch.size()][];
IntStream.range(0, this.batch.size()).forEach(x -> predictionMatrix[x] = this.batch.get(x).getPrediction(threshold));
return predictionMatrix;
}
@Override
public int[][] getThresholdedPredictionMatrix(final double[] threshold) {
int[][] predictionMatrix = new int[this.batch.size()][];
IntStream.range(0, this.batch.size()).forEach(x -> predictionMatrix[x] = this.batch.get(x).getPrediction(threshold));
return predictionMatrix;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/package-info.java
|
package ai.libs.jaicore.ml.classification.multilabel;
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/AMultiLabelClassificationMeasure.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss;
import java.util.Arrays;
import java.util.List;
import java.util.Set;
import java.util.stream.Collectors;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import org.api4.java.ai.ml.classification.multilabel.evaluation.loss.IMultiLabelClassificationPredictionPerformanceMeasure;
import ai.libs.jaicore.ml.classification.loss.dataset.APredictionPerformanceMeasure;
public abstract class AMultiLabelClassificationMeasure extends APredictionPerformanceMeasure<int[], IMultiLabelClassification> implements IMultiLabelClassificationPredictionPerformanceMeasure {
private static final double DEFAULT_THRESHOLD = 0.5;
private final double threshold;
protected AMultiLabelClassificationMeasure(final double threshold) {
super();
this.threshold = threshold;
}
protected AMultiLabelClassificationMeasure() {
this(DEFAULT_THRESHOLD);
}
public double getThreshold() {
return this.threshold;
}
protected double[][] listToRelevanceMatrix(final List<? extends IMultiLabelClassification> classificationList) {
double[][] matrix = new double[classificationList.size()][];
IntStream.range(0, classificationList.size()).forEach(x -> matrix[x] = classificationList.get(x).getPrediction());
return matrix;
}
protected int[][] listToThresholdedRelevanceMatrix(final List<? extends IMultiLabelClassification> classificationList) {
int[][] matrix = new int[classificationList.size()][];
IntStream.range(0, classificationList.size()).forEach(x -> matrix[x] = classificationList.get(x).getPrediction(this.threshold));
return matrix;
}
protected Set<Integer> getThresholdedPredictionAsSet(final IMultiLabelClassification prediction) {
return Arrays.stream(prediction.getThresholdedPrediction()).mapToObj(Integer::valueOf).collect(Collectors.toSet());
}
protected int[][] listToMatrix(final List<? extends int[]> classificationList) {
int[][] matrix = new int[classificationList.size()][];
IntStream.range(0, classificationList.size()).forEach(x -> matrix[x] = classificationList.get(x));
return matrix;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/AThresholdBasedMultiLabelClassificationMeasure.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss;
import java.util.List;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import org.api4.java.ai.ml.classification.multilabel.evaluation.loss.IMultiLabelClassificationPredictionPerformanceMeasure;
import ai.libs.jaicore.ml.classification.loss.dataset.APredictionPerformanceMeasure;
public abstract class AThresholdBasedMultiLabelClassificationMeasure extends APredictionPerformanceMeasure<int[], IMultiLabelClassification> implements IMultiLabelClassificationPredictionPerformanceMeasure {
private static final double DEFAULT_THRESHOLD = 0.5;
private final double threshold;
protected AThresholdBasedMultiLabelClassificationMeasure(final double threshold) {
super();
this.threshold = threshold;
}
protected AThresholdBasedMultiLabelClassificationMeasure() {
super();
this.threshold = DEFAULT_THRESHOLD;
}
public double getThreshold() {
return this.threshold;
}
protected double[][] listToRelevanceMatrix(final List<? extends IMultiLabelClassification> classificationList) {
double[][] matrix = new double[classificationList.size()][];
IntStream.range(0, classificationList.size()).forEach(x -> matrix[x] = classificationList.get(x).getPrediction());
return matrix;
}
protected int[][] listToThresholdedRelevanceMatrix(final List<? extends IMultiLabelClassification> classificationList) {
int[][] matrix = new int[classificationList.size()][];
IntStream.range(0, classificationList.size()).forEach(x -> matrix[x] = classificationList.get(x).getPrediction(this.threshold));
return matrix;
}
protected int[][] listToMatrix(final List<? extends int[]> classificationList) {
int[][] matrix = new int[classificationList.size()][];
IntStream.range(0, classificationList.size()).forEach(x -> matrix[x] = classificationList.get(x));
return matrix;
}
protected int[][] transposeMatrix(final int[][] matrix) {
int[][] out = new int[matrix[0].length][];
for (int i = 0; i < matrix[0].length; i++) {
out[i] = new int[matrix.length];
for (int j = 0; j < matrix.length; j++) {
out[i][j] = matrix[j][i];
}
}
return out;
}
protected double[][] transposeMatrix(final double[][] matrix) {
double[][] out = new double[matrix[0].length][];
for (int i = 0; i < matrix[0].length; i++) {
out[i] = new double[matrix.length];
for (int j = 0; j < matrix.length; j++) {
out[i][j] = matrix[j][i];
}
}
return out;
}
@Override
public int hashCode() {
final int prime = 31;
int result = 1;
long temp;
temp = Double.doubleToLongBits(this.threshold);
result = prime * result + (int) (temp ^ (temp >>> 32));
return result;
}
@Override
public boolean equals(final Object obj) {
if (this == obj) {
return true;
}
if (obj == null || this.getClass() != obj.getClass()) {
return false;
}
return Double.doubleToLongBits(this.threshold) == Double.doubleToLongBits(((AThresholdBasedMultiLabelClassificationMeasure) obj).threshold);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/AutoMEKAGGPFitnessMeasureLoss.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss;
import java.util.Arrays;
import java.util.List;
import java.util.OptionalDouble;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import org.api4.java.ai.ml.classification.multilabel.evaluation.loss.IMultiLabelClassificationPredictionPerformanceMeasure;
/**
* Measure combining exact match, hamming loss, f1macroavgL and rankloss. Here
* implemented in inverse.
*
* de Sa, Alex GC, Gisele L. Pappa, and Alex A. Freitas. "Towards a method for automatically selecting and configuring multi-label
* classification algorithms." Proceedings of the Genetic and Evolutionary Computation Conference Companion. ACM, 2017.
*
* @author mwever
*
*/
public class AutoMEKAGGPFitnessMeasureLoss extends AMultiLabelClassificationMeasure {
private IMultiLabelClassificationPredictionPerformanceMeasure[] measures;
@SuppressWarnings("unchecked")
public AutoMEKAGGPFitnessMeasureLoss() {
super();
this.measures = new IMultiLabelClassificationPredictionPerformanceMeasure[] { new ExactMatch(), new F1MacroAverageL(), new Hamming(), new RankLoss() };
}
@SuppressWarnings("unchecked")
public AutoMEKAGGPFitnessMeasureLoss(final double threshold) {
super(threshold);
this.measures = new IMultiLabelClassificationPredictionPerformanceMeasure[] { new ExactMatch(threshold), new F1MacroAverageL(threshold), new Hamming(threshold), new RankLoss(threshold) };
}
@Override
public double loss(final List<? extends int[]> expected, final List<? extends IMultiLabelClassification> predicted) {
OptionalDouble res = Arrays.stream(this.measures).mapToDouble(x -> x.loss(expected, predicted)).average();
if (res.isPresent()) {
return res.getAsDouble();
} else {
throw new IllegalStateException("Could not take the average of all base measures");
}
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/ExactMatch.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss;
import java.util.List;
import java.util.stream.IntStream;
import org.apache.commons.collections4.SetUtils;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import ai.libs.jaicore.basic.ArrayUtil;
public class ExactMatch extends AMultiLabelClassificationMeasure {
public ExactMatch() {
super();
}
public ExactMatch(final double threshold) {
super(threshold);
}
@Override
public double loss(final List<? extends int[]> expected, final List<? extends IMultiLabelClassification> predicted) {
this.checkConsistency(expected, predicted);
return (double) IntStream.range(0, expected.size()).map(x -> SetUtils.isEqualSet(ArrayUtil.argMax(expected.get(x)), this.getThresholdedPredictionAsSet(predicted.get(x))) ? 0 : 1).sum() / expected.size();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/F1MacroAverageL.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss;
import java.util.Arrays;
import java.util.List;
import java.util.OptionalDouble;
import java.util.stream.Collectors;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import ai.libs.jaicore.basic.ArrayUtil;
import ai.libs.jaicore.ml.classification.loss.dataset.F1Measure;
import ai.libs.jaicore.ml.classification.singlelabel.SingleLabelClassification;
public class F1MacroAverageL extends AMultiLabelClassificationMeasure {
public F1MacroAverageL(final double threshold) {
super(threshold);
}
public F1MacroAverageL() {
super();
}
@Override
public double score(final List<? extends int[]> expected, final List<? extends IMultiLabelClassification> predicted) {
this.checkConsistency(expected, predicted);
int[][] expectedMatrix = ArrayUtil.transposeMatrix(this.listToMatrix(expected));
int[][] actualMatrix = ArrayUtil.transposeMatrix(this.listToThresholdedRelevanceMatrix(predicted));
F1Measure loss = new F1Measure(1);
OptionalDouble res = IntStream.range(0, expectedMatrix.length).mapToDouble(
x -> loss.score(Arrays.stream(expectedMatrix[x]).mapToObj(Integer::valueOf).collect(Collectors.toList()), Arrays.stream(actualMatrix[x]).mapToObj(y -> new SingleLabelClassification(2, y)).collect(Collectors.toList())))
.average();
if (!res.isPresent()) {
throw new IllegalStateException("Could not determine average label-wise f measure.");
} else {
return res.getAsDouble();
}
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/F1MicroAverage.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss;
import java.util.Arrays;
import java.util.List;
import java.util.stream.Collectors;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
import ai.libs.jaicore.ml.classification.loss.dataset.F1Measure;
import ai.libs.jaicore.ml.classification.singlelabel.SingleLabelClassification;
public class F1MicroAverage extends AMultiLabelClassificationMeasure {
public F1MicroAverage(final double threshold) {
super(threshold);
}
public F1MicroAverage() {
super();
}
@Override
public double score(final List<? extends int[]> expected, final List<? extends IMultiLabelClassification> predicted) {
this.checkConsistency(expected, predicted);
List<Integer> expectedMatrix = expected.stream().flatMapToInt(Arrays::stream).mapToObj(x -> x).collect(Collectors.toList());
List<ISingleLabelClassification> predictedMatrix = expected.stream().flatMapToInt(Arrays::stream).mapToObj(x -> new SingleLabelClassification(2, x)).collect(Collectors.toList());
F1Measure loss = new F1Measure(1);
return loss.score(expectedMatrix, predictedMatrix);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/Hamming.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss;
import java.util.List;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import ai.libs.jaicore.basic.ArrayUtil;
import ai.libs.jaicore.basic.sets.SetUtil;
public class Hamming extends AMultiLabelClassificationMeasure {
public Hamming() {
super();
}
public Hamming(final double threshold) {
super(threshold);
}
@Override
public double loss(final List<? extends int[]> expected, final List<? extends IMultiLabelClassification> predicted) {
this.checkConsistency(expected, predicted);
return (double) IntStream.range(0, expected.size()).map(x -> SetUtil.getDisjointSet(this.getThresholdedPredictionAsSet(predicted.get(x)), ArrayUtil.argMax(expected.get(x))).size()).sum() / (expected.size() * expected.get(0).length);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/InstanceWiseF1.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss;
import java.util.Arrays;
import java.util.List;
import java.util.OptionalDouble;
import java.util.stream.Collectors;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import ai.libs.jaicore.ml.classification.loss.dataset.F1Measure;
import ai.libs.jaicore.ml.classification.singlelabel.SingleLabelClassification;
/**
* Instance-wise F1 measure for multi-label classifiers.
*
* For reference see
* Wu, Xi-Zhu; Zhou, Zhi-Hua: A Unified View of Multi-Label Performance Measures (ICML / JMLR 2017)
*
* @author mwever
*
*/
public class InstanceWiseF1 extends AMultiLabelClassificationMeasure {
public InstanceWiseF1(final double threshold) {
super(threshold);
}
public InstanceWiseF1() {
super();
}
@Override
public double score(final List<? extends int[]> expected, final List<? extends IMultiLabelClassification> predicted) {
this.checkConsistency(expected, predicted);
int[][] expectedMatrix = this.listToMatrix(expected);
int[][] actualMatrix = this.listToThresholdedRelevanceMatrix(predicted);
F1Measure baseMeasure = new F1Measure(1);
OptionalDouble res = IntStream.range(0, expectedMatrix.length)
.mapToDouble(
x -> baseMeasure.score(Arrays.stream(expectedMatrix[x]).mapToObj(y -> y).collect(Collectors.toList()), Arrays.stream(actualMatrix[x]).mapToObj(y -> new SingleLabelClassification(2, y)).collect(Collectors.toList())))
.average();
if (!res.isPresent() || Double.isNaN(res.getAsDouble())) {
throw new IllegalStateException("Could not determine average instance-wise f measure.");
} else {
return res.getAsDouble();
}
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/JaccardScore.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss;
import java.util.List;
import java.util.OptionalDouble;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import ai.libs.jaicore.basic.ArrayUtil;
public class JaccardScore extends AMultiLabelClassificationMeasure {
private ai.libs.jaicore.ml.classification.loss.instance.JaccardScore instanceScorer;
public JaccardScore() {
super();
}
public JaccardScore(final double threshold) {
super(threshold);
}
@Override
public double score(final List<? extends int[]> expected, final List<? extends IMultiLabelClassification> predicted) {
OptionalDouble res = IntStream.range(0, expected.size()).mapToDouble(x -> this.instanceScorer.score(ArrayUtil.argMax(expected.get(x)), this.getThresholdedPredictionAsSet(predicted.get(x)))).average();
if (!res.isPresent()) {
throw new IllegalStateException("Could not average the jaccord score.");
} else {
return res.getAsDouble();
}
}
@Override
public double loss(final List<? extends int[]> expected, final List<? extends IMultiLabelClassification> predicted) {
return 1 - this.score(expected, predicted);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/RankLoss.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss;
import java.util.List;
import java.util.OptionalDouble;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import ai.libs.jaicore.basic.ArrayUtil;
public class RankLoss extends AMultiLabelClassificationMeasure {
private static final double DEFAULT_TIE_LOSS = 0.0;
private final double tieLoss;
public RankLoss() {
this(DEFAULT_TIE_LOSS);
}
/**
* Create a Ranking Loss measure instance.
*
* @param tieLoss The loss [0,1] which is accounted for a tie of the predicted relevant and irrelevant label's relevance score.
*/
public RankLoss(final double tieLoss) {
this.tieLoss = tieLoss;
}
private double rankingLoss(final int[] expected, final IMultiLabelClassification predicted) {
List<Integer> expectedRelevantLabels = ArrayUtil.argMax(expected);
List<Integer> expectedIrrelevantLabels = ArrayUtil.argMin(expected);
double[] labelRelevance = predicted.getPrediction();
double wrongRankingCounter = 0;
for (int expectedRel : expectedRelevantLabels) {
for (int expectedIrr : expectedIrrelevantLabels) {
double scoreRelLabel = labelRelevance[expectedRel];
double scoreIrrLabel = labelRelevance[expectedIrr];
if (scoreRelLabel == scoreIrrLabel) {
wrongRankingCounter += this.tieLoss;
} else if (scoreRelLabel < scoreIrrLabel) {
wrongRankingCounter += 1.0;
}
}
}
return wrongRankingCounter / (expectedRelevantLabels.size() + expectedIrrelevantLabels.size());
}
@Override
public double loss(final List<? extends int[]> expected, final List<? extends IMultiLabelClassification> predicted) {
this.checkConsistency(expected, predicted);
OptionalDouble res = IntStream.range(0, expected.size()).mapToDouble(x -> this.rankingLoss(expected.get(x), predicted.get(x))).average();
if (res.isPresent()) {
return res.getAsDouble();
} else {
throw new IllegalStateException("The ranking loss could not be averaged across all the instances.");
}
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/package-info.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss;
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive/ChoquisticRelevanceLoss.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.nonadditive;
import java.util.Collections;
import java.util.List;
import java.util.stream.Collectors;
import java.util.stream.IntStream;
import org.apache.commons.math3.stat.descriptive.DescriptiveStatistics;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.AMultiLabelClassificationMeasure;
import ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.nonadditive.choquistic.IMassFunction;
public class ChoquisticRelevanceLoss extends AMultiLabelClassificationMeasure {
private final IMassFunction measure;
public ChoquisticRelevanceLoss(final IMassFunction measure) {
super();
this.measure = measure;
}
public ChoquisticRelevanceLoss(final double threshold, final IMassFunction measure) {
super(threshold);
this.measure = measure;
}
/**
* Function f to be integrated: f(c_i) = u_i = 1 - | s_i - y_i |
* s_i \in [0,1]: label relevance score predicted
* y_i \in {0,1}: ground truth
*
* @param expected
* @param predicted
* @return
*/
private double fcrit(final int expected, final double predicted) {
return 1 - Math.abs(predicted - expected);
}
private double instanceLoss(final int[] expected, final double[] predicted) {
double sum = 0.0;
List<Double> listOfCis = IntStream.range(0, expected.length).mapToObj(i -> this.fcrit(expected[i], predicted[i])).collect(Collectors.toList());
listOfCis.add(0.0);
Collections.sort(listOfCis);
for (int i = 1; i < listOfCis.size(); i++) {
sum += (listOfCis.get(i) - listOfCis.get(i - 1)) * this.measure.mu(IntStream.range(i, listOfCis.size()).mapToObj(listOfCis::get).collect(Collectors.toList()), expected.length);
}
return sum;
}
@Override
public double loss(final List<? extends int[]> expected, final List<? extends IMultiLabelClassification> predicted) {
this.checkConsistency(expected, predicted);
DescriptiveStatistics stats = new DescriptiveStatistics();
for (int i = 0; i < expected.size(); i++) {
stats.addValue(this.instanceLoss(expected.get(i), predicted.get(i).getPrediction()));
}
return stats.getMean();
}
@Override
public double score(final List<? extends int[]> expected, final List<? extends IMultiLabelClassification> predicted) {
return 1 - this.loss(expected, predicted);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive/OWARelevanceLoss.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.nonadditive;
import java.util.Collections;
import java.util.List;
import java.util.stream.Collectors;
import java.util.stream.IntStream;
import org.apache.commons.math3.stat.descriptive.DescriptiveStatistics;
import org.api4.java.ai.ml.classification.multilabel.evaluation.IMultiLabelClassification;
import ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.AMultiLabelClassificationMeasure;
import ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.nonadditive.owa.IOWAValueFunction;
public class OWARelevanceLoss extends AMultiLabelClassificationMeasure {
private final IOWAValueFunction valueFunction;
public OWARelevanceLoss(final IOWAValueFunction valueFunction) {
this.valueFunction = valueFunction;
}
/**
* Function f to be integrated: f(c_i) = u_i = 1 - | s_i - y_i |
* s_i \in [0,1]: label relevance score predicted
* y_i \in {0,1}: ground truth
*
* @param expected
* @param predicted
* @return
*/
private double fcrit(final int expected, final double predicted) {
return 1 - Math.abs(predicted - expected);
}
private double instanceLoss(final int[] expected, final double[] predicted) {
double sum = 0.0;
double m = expected.length;
List<Double> listOfCis = IntStream.range(0, expected.length).mapToObj(i -> this.fcrit(expected[i], predicted[i])).collect(Collectors.toList());
listOfCis.add(0.0);
Collections.sort(listOfCis);
for (int i = 1; i < listOfCis.size(); i++) {
sum += (this.valueFunction.transform((m - i + 1), m) - this.valueFunction.transform((m - i), m)) * listOfCis.get(i);
}
return 1 - sum;
}
@Override
public double loss(final List<? extends int[]> expected, final List<? extends IMultiLabelClassification> predicted) {
this.checkConsistency(expected, predicted);
DescriptiveStatistics stats = new DescriptiveStatistics();
for (int i = 0; i < expected.size(); i++) {
stats.addValue(this.instanceLoss(expected.get(i), predicted.get(i).getPrediction()));
}
return stats.getMean();
}
@Override
public double score(final List<? extends int[]> expected, final List<? extends IMultiLabelClassification> predicted) {
return 1 - this.loss(expected, predicted);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive/package-info.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.nonadditive;
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive/choquistic/HammingMassFunction.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.nonadditive.choquistic;
import java.util.Collection;
public class HammingMassFunction implements IMassFunction {
@Override
public double mu(final Collection<Double> cis, final int m) {
return (double) cis.size() / m;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive/choquistic/IMassFunction.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.nonadditive.choquistic;
import java.util.Collection;
public interface IMassFunction {
public double mu(final Collection<Double> cis, final int m);
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive/choquistic/SubsetZeroOneMassFunction.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.nonadditive.choquistic;
import java.util.Collection;
public class SubsetZeroOneMassFunction implements IMassFunction {
@Override
public double mu(final Collection<Double> cis, final int m) {
if (cis.size() == m) {
return 1;
} else {
return 0;
}
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive/choquistic/package-info.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.nonadditive.choquistic;
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive/owa/IOWAValueFunction.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.nonadditive.owa;
public interface IOWAValueFunction {
public double transform(double nominator, double denominator);
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive/owa/MoebiusTransformOWAValueFunction.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.nonadditive.owa;
import org.apache.commons.math3.util.CombinatoricsUtils;
public class MoebiusTransformOWAValueFunction implements IOWAValueFunction {
private final int k;
public MoebiusTransformOWAValueFunction(final int k) {
this.k = k;
}
@Override
public double transform(final double nominator, final double denominator) {
if ((int) nominator >= this.k) {
return CombinatoricsUtils.binomialCoefficientDouble((int) nominator, this.k) / CombinatoricsUtils.binomialCoefficientDouble((int) denominator, this.k);
} else {
return 0;
}
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive/owa/PolynomialOWAValueFunction.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.nonadditive.owa;
public class PolynomialOWAValueFunction implements IOWAValueFunction {
private final double alpha;
public PolynomialOWAValueFunction(final double alpha) {
this.alpha = alpha;
}
@Override
public double transform(final double nominator, final double denominator) {
return Math.pow(nominator / denominator, this.alpha);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/multilabel/evaluation/loss/nonadditive/owa/package-info.java
|
package ai.libs.jaicore.ml.classification.multilabel.evaluation.loss.nonadditive.owa;
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/SingleLabelClassification.java
|
package ai.libs.jaicore.ml.classification.singlelabel;
import java.util.Arrays;
import java.util.HashMap;
import java.util.Map;
import java.util.Map.Entry;
import java.util.stream.IntStream;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
import ai.libs.jaicore.basic.ArrayUtil;
import ai.libs.jaicore.ml.core.evaluation.Prediction;
public class SingleLabelClassification extends Prediction implements ISingleLabelClassification {
private double[] labelProbabilities;
public SingleLabelClassification(final int numClasses, final int predicted) {
super(predicted);
this.labelProbabilities = new double[numClasses];
this.labelProbabilities[predicted] = 1.0;
}
public SingleLabelClassification(final Map<Integer, Double> labelProbabilities) {
super(labelWithHighestProbability(labelProbabilities));
this.labelProbabilities = new double[labelProbabilities.size()];
labelProbabilities.entrySet().stream().forEach(x -> this.labelProbabilities[x.getKey()] = x.getValue());
}
public SingleLabelClassification(final double[] labelProbabilities) {
super(ArrayUtil.argMax(labelProbabilities).get(0));
this.labelProbabilities = labelProbabilities;
}
@Override
public int getIntPrediction() {
return (int) super.getPrediction();
}
@Override
public Integer getPrediction() {
return this.getIntPrediction();
}
@Override
public Integer getLabelWithHighestProbability() {
return this.getIntPrediction();
}
@Override
public Map<Integer, Double> getClassDistribution() {
Map<Integer, Double> distributionMap = new HashMap<>();
IntStream.range(0, this.labelProbabilities.length).forEach(x -> distributionMap.put(x, this.labelProbabilities[x]));
return distributionMap;
}
@Override
public double getProbabilityOfLabel(final int label) {
return this.labelProbabilities[label];
}
@Override
public Map<Integer, Double> getClassConfidence() {
Map<Integer, Double> confidenceMap = new HashMap<>();
IntStream.range(0, this.labelProbabilities.length).forEach(x -> confidenceMap.put(x, this.labelProbabilities[x]));
return confidenceMap;
}
private static int labelWithHighestProbability(final Map<Integer, Double> labelProbabilities) {
Entry<Integer, Double> highestProbEntry = null;
for (Entry<Integer, Double> entry : labelProbabilities.entrySet()) {
if (highestProbEntry == null || highestProbEntry.getValue() < entry.getValue()) {
highestProbEntry = entry;
}
}
if (highestProbEntry == null) {
throw new IllegalArgumentException("No prediction contained");
} else {
return highestProbEntry.getKey();
}
}
@Override
public String toString() {
return new StringBuilder().append(this.getPrediction()).append(" ").append(Arrays.toString(this.labelProbabilities)).toString();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/SingleLabelClassificationPredictionBatch.java
|
package ai.libs.jaicore.ml.classification.singlelabel;
import java.util.ArrayList;
import java.util.Collection;
import java.util.List;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassificationPredictionBatch;
public class SingleLabelClassificationPredictionBatch extends ArrayList<ISingleLabelClassification> implements ISingleLabelClassificationPredictionBatch {
/**
*
*/
private static final long serialVersionUID = 3575940001172802462L;
public SingleLabelClassificationPredictionBatch(final Collection<ISingleLabelClassification> predictions) {
this.addAll(predictions);
}
@Override
public int getNumPredictions() {
return this.size();
}
@Override
public List<? extends ISingleLabelClassification> getPredictions() {
return this;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/package-info.java
|
package ai.libs.jaicore.ml.classification.singlelabel;
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/learner/ASingleLabelClassifier.java
|
package ai.libs.jaicore.ml.classification.singlelabel.learner;
import java.util.LinkedList;
import java.util.List;
import java.util.Map;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassificationPredictionBatch;
import org.api4.java.ai.ml.core.dataset.supervised.ILabeledDataset;
import org.api4.java.ai.ml.core.dataset.supervised.ILabeledInstance;
import org.api4.java.ai.ml.core.exception.PredictionException;
import ai.libs.jaicore.ml.classification.singlelabel.SingleLabelClassificationPredictionBatch;
import ai.libs.jaicore.ml.core.learner.ASupervisedLearner;
public abstract class ASingleLabelClassifier extends ASupervisedLearner<ILabeledInstance, ILabeledDataset<? extends ILabeledInstance>, ISingleLabelClassification, ISingleLabelClassificationPredictionBatch> {
protected ASingleLabelClassifier(final Map<String, Object> config) {
super(config);
}
protected ASingleLabelClassifier() {
super();
}
@Override
public ISingleLabelClassificationPredictionBatch predict(final ILabeledInstance[] dTest) throws PredictionException, InterruptedException {
List<ISingleLabelClassification> batchList = new LinkedList<>();
for (ILabeledInstance i : dTest) {
batchList.add(this.predict(i));
}
return new SingleLabelClassificationPredictionBatch(batchList);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/learner/MajorityClassifier.java
|
package ai.libs.jaicore.ml.classification.singlelabel.learner;
import java.util.HashMap;
import java.util.Map;
import java.util.Objects;
import org.api4.java.ai.ml.classification.IClassifier;
import org.api4.java.ai.ml.core.dataset.schema.attribute.ICategoricalAttribute;
import org.api4.java.ai.ml.core.dataset.supervised.ILabeledDataset;
import org.api4.java.ai.ml.core.dataset.supervised.ILabeledInstance;
import org.api4.java.ai.ml.core.evaluation.IPrediction;
import org.api4.java.ai.ml.core.evaluation.IPredictionBatch;
import org.api4.java.ai.ml.core.exception.PredictionException;
import org.api4.java.ai.ml.core.exception.TrainingException;
import ai.libs.jaicore.ml.classification.singlelabel.SingleLabelClassification;
import ai.libs.jaicore.ml.core.evaluation.PredictionBatch;
import ai.libs.jaicore.ml.core.learner.ASupervisedLearner;
public class MajorityClassifier extends ASupervisedLearner<ILabeledInstance, ILabeledDataset<? extends ILabeledInstance>, IPrediction, IPredictionBatch> implements IClassifier {
private Object majorityLabel;
private double[] prediction;
@Override
public void fit(final ILabeledDataset<? extends ILabeledInstance> dTrain) throws TrainingException, InterruptedException {
if (!(dTrain.getLabelAttribute() instanceof ICategoricalAttribute)) {
throw new IllegalArgumentException("The label attribute of the given data is of type " + dTrain.getLabelAttribute().getClass() + ", but the " + MajorityClassifier.class.getName() + " can only work with categorical labels.");
}
Map<Object, Integer> labelCounter = new HashMap<>();
Objects.requireNonNull(dTrain);
if (dTrain.isEmpty()) {
throw new IllegalArgumentException("Cannot train majority classifier with empty training set.");
}
for (ILabeledInstance i : dTrain) {
labelCounter.put(i.getLabel(), labelCounter.computeIfAbsent(i.getLabel(), t -> 0) + 1);
}
this.majorityLabel = labelCounter.keySet().stream().max((l1, l2) -> Integer.compare(labelCounter.get(l1), labelCounter.get(l2))).get();
ICategoricalAttribute labelAtt = ((ICategoricalAttribute) dTrain.getLabelAttribute());
this.prediction = new double[labelAtt.getLabels().size()];
this.prediction[labelAtt.getAsAttributeValue(this.majorityLabel).getValue()] = 1.0;
}
@Override
public IPrediction predict(final ILabeledInstance xTest) throws PredictionException, InterruptedException {
return new SingleLabelClassification(this.prediction);
}
@Override
public IPredictionBatch predict(final ILabeledInstance[] dTest) throws PredictionException, InterruptedException {
IPrediction[] predictions = new IPrediction[dTest.length];
for (int i = 0; i < dTest.length; i++) {
predictions[i] = this.predict(dTest[i]);
}
return new PredictionBatch(predictions);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/learner/package-info.java
|
package ai.libs.jaicore.ml.classification.singlelabel.learner;
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/package-info.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries;
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset/ITimeSeriesDataset.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset;
import java.util.Iterator;
public interface ITimeSeriesDataset {
public Iterator<ITimeSeriesInstance> iterator();
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset/ITimeSeriesInstance.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset;
import org.api4.java.ai.ml.core.dataset.supervised.ILabeledInstance;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.model.INDArrayTimeseries;
import ai.libs.jaicore.ml.core.filter.sampling.IClusterableInstance;
public interface ITimeSeriesInstance extends IClusterableInstance, ILabeledInstance, Iterable<INDArrayTimeseries> {
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset/TimeSeriesDataset.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset;
import java.util.ArrayList;
import java.util.Iterator;
import java.util.List;
import java.util.NoSuchElementException;
import java.util.Set;
import java.util.stream.Collectors;
import org.api4.java.ai.ml.core.dataset.schema.ILabeledInstanceSchema;
import org.api4.java.ai.ml.core.dataset.schema.attribute.IAttribute;
import org.api4.java.ai.ml.core.dataset.schema.attribute.ITimeseriesAttribute;
import org.api4.java.ai.ml.core.dataset.supervised.ILabeledDataset;
import org.api4.java.ai.ml.core.exception.DatasetCreationException;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.factory.Nd4j;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.attribute.NDArrayTimeseriesAttribute;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.model.INDArrayTimeseries;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.model.NDArrayTimeseries;
import ai.libs.jaicore.ml.core.dataset.ADataset;
/**
* Time Series Dataset.
*/
public class TimeSeriesDataset extends ADataset<ITimeSeriesInstance> implements ILabeledDataset<ITimeSeriesInstance> {
private static final long serialVersionUID = -6819487387561457394L;
/** Values of time series variables. */
private List<INDArray> valueMatrices;
/** Timestamps of time series variables. */
private List<INDArray> timestampMatrices;
/** Target values for the instances. */
private transient List<Object> targets;
/**
* Creates a TimeSeries dataset. Let `n` be the number of instances.
*
* @param valueMatrices Values for the time series variables. List of
* 2D-Arrays with shape `[n, ?]`.
* @param timestampMatrices Timestamps for the time series variables. List of
* 2D-Arrays with shape `[n, ?]`. Or `null` if no
* timestamps exist for the corresponding time series
* variable. The shape of the `i`th index must be equal
* to the shape of the `i`th element of
* `valueMatrices`.
* @param targets Target values for the instances.
*/
public TimeSeriesDataset(final ILabeledInstanceSchema schema, final List<INDArray> valueMatrices, final List<INDArray> timestampMatrices, final List<Object> targets) {
this(schema);
for (IAttribute att : schema.getAttributeList()) {
if (!(att instanceof ITimeseriesAttribute)) {
throw new IllegalArgumentException("The schema contains attributes which are not timeseries");
}
}
Set<Object> valueInstances = valueMatrices.stream().map(x -> x.shape()[0]).collect(Collectors.toSet());
if (valueInstances.size() > 1) {
throw new IllegalArgumentException("The value matrices vary in length i.e. they have different number of instances");
}
Set<Object> timestampInstances = timestampMatrices.stream().map(x -> x.shape()[0]).collect(Collectors.toSet());
if (timestampInstances.size() > 1) {
throw new IllegalArgumentException("The timestamp matrices vary in length i.e. they have different number of instances");
}
valueInstances.addAll(timestampInstances);
if (valueInstances.size() > 1) {
throw new IllegalArgumentException("There are different number of instances for values and timestamps");
}
this.valueMatrices = valueMatrices;
this.timestampMatrices = timestampMatrices;
this.targets = targets;
}
public TimeSeriesDataset(final ILabeledInstanceSchema schema) {
super(schema);
}
/**
* Add a time series variable to the dataset.
*
* @param valueMatrix Values for the time series variable to add. 2D-Arrays
* with shape `[n, ?]` where `n` is the number of
* instances of the dataset.
* @param timestampMatrix Timestamps for the time series variable to add.
* 2D-Arrays with shape `[n, ?]` where `n` is the number
* of instances of the dataset. Or `null` if no timestamp
* exists for this time series variable.
*/
public void add(final String attributeName, final INDArray valueMatrix, final INDArray timestampMatrix) {
// Parameter checks.
// ..
this.valueMatrices.add(valueMatrix);
this.timestampMatrices.add(timestampMatrix);
this.addAttribute(attributeName, valueMatrix);
}
/**
* Removes the time series variable at a given index.
*
* @param index
* @return
* @throws IndexOutOfBoundsException
*/
@Override
public void removeColumn(final int index) {
this.valueMatrices.remove(index);
this.timestampMatrices.remove(index);
this.getInstanceSchema().removeAttribute(index);
}
/**
* Replaces the time series variable at a given index with a new one.
*
* @param index Index of the time series variable to replace.
* @param valueMatrix Values for the time series variable to add. 2D-Arrays
* with shape `[n, ?]` where `n` is the number of
* instances of the dataset.
* @param timestampMatrix Timestamps for the time series variable to add.
* 2D-Arrays with shape `[n, ?]` where `n` is the number
* of instances of the dataset. Or `null` if no timestamp
* exists for this time series variable.
* @throws IndexOutOfBoundsException Thrown if `numberOfInstances <= index`.
*/
public void replace(final int index, final INDArray valueMatrix, final INDArray timestampMatrix) {
this.valueMatrices.set(index, valueMatrix);
if (timestampMatrix != null && this.timestampMatrices != null && this.timestampMatrices.size() > index) {
this.timestampMatrices.set(index, timestampMatrix);
}
NDArrayTimeseriesAttribute type = this.createAttribute("ts" + index, valueMatrix);
this.getInstanceSchema().removeAttribute(index);
this.getInstanceSchema().addAttribute(index, type);
}
public Object getTargets() {
return this.targets;
}
public INDArray getTargetsAsINDArray() {
if (this.targets.get(0) instanceof Number) {
return Nd4j.create(this.targets.stream().mapToDouble(x -> (Double) x).toArray());
}
return null;
}
public int getNumberOfVariables() {
return this.valueMatrices.size();
}
public long getNumberOfInstances() {
return this.valueMatrices.get(0).shape()[0];
}
public INDArray getValues(final int index) {
return this.valueMatrices.get(index);
}
public INDArray getTimestamps(final int index) {
return this.timestampMatrices.get(index);
}
public INDArray getValuesOrNull(final int index) {
return this.valueMatrices.size() > index ? this.valueMatrices.get(index) : null;
}
public INDArray getTimestampsOrNull(final int index) {
return this.timestampMatrices != null && this.timestampMatrices.size() > index ? this.timestampMatrices.get(index) : null;
}
@Override
public boolean isEmpty() {
return this.valueMatrices.isEmpty();
}
public boolean isUnivariate() {
return this.valueMatrices.size() == 1;
}
public boolean isMultivariate() {
return this.valueMatrices.size() > 1;
}
// --
// Intern helper functions.
// --
private NDArrayTimeseriesAttribute createAttribute(final String name, final INDArray valueMatrix) {
int length = (int) valueMatrix.shape()[1];
return new NDArrayTimeseriesAttribute(name, length);
}
private void addAttribute(final String name, final INDArray valueMatrix) {
NDArrayTimeseriesAttribute type = this.createAttribute(name, valueMatrix);
this.getInstanceSchema().addAttribute(type);
this.valueMatrices.add(valueMatrix);
}
// --
// IDataset interface.
// --
/**
* Iterator for the @{@link}TimeSeriesDataset. Iterates and implicitly creates
* the @{link}TimeSeriesInstance.
*/
class TimeSeriesDatasetIterator implements Iterator<ITimeSeriesInstance> {
private int current = 0;
@Override
public boolean hasNext() {
return TimeSeriesDataset.this.getNumberOfInstances() > this.current;
}
@Override
public ITimeSeriesInstance next() {
if (!this.hasNext()) {
throw new NoSuchElementException();
}
return TimeSeriesDataset.this.get(this.current++);
}
}
@Override
public TimeSeriesInstance get(final int index) {
// Build attribute value as view on the row of the attribute matrix.
List<INDArrayTimeseries> attributeValues = new ArrayList<>();
for (int i = 0; i < TimeSeriesDataset.this.valueMatrices.size(); i++) {
attributeValues.add(new NDArrayTimeseries(TimeSeriesDataset.this.valueMatrices.get(i).getRow(index)));
}
// Build target value.
Object target = this.targets.get(index);
return new TimeSeriesInstance(attributeValues, target);
}
@Override
public Iterator<ITimeSeriesInstance> iterator() {
return new TimeSeriesDatasetIterator();
}
@Override
public TimeSeriesDataset createEmptyCopy() throws DatasetCreationException, InterruptedException {
return new TimeSeriesDataset(this.getInstanceSchema());
}
@Override
public Object[][] getFeatureMatrix() {
throw new UnsupportedOperationException();
}
@Override
public Object[] getLabelVector() {
return this.targets.toArray();
}
@Override
public TimeSeriesDataset createCopy() throws DatasetCreationException, InterruptedException {
TimeSeriesDataset copy = this.createEmptyCopy();
for (ITimeSeriesInstance i : this) {
copy.add(i);
}
return copy;
}
@Override
public int hashCode() {
final int prime = 31;
int result = super.hashCode();
result = prime * result + ((this.targets == null) ? 0 : this.targets.hashCode());
result = prime * result + ((this.timestampMatrices == null) ? 0 : this.timestampMatrices.hashCode());
result = prime * result + ((this.valueMatrices == null) ? 0 : this.valueMatrices.hashCode());
return result;
}
@Override
public boolean equals(final Object obj) {
if (this == obj) {
return true;
}
if (!super.equals(obj)) {
return false;
}
if (this.getClass() != obj.getClass()) {
return false;
}
TimeSeriesDataset other = (TimeSeriesDataset) obj;
if (this.targets == null) {
if (other.targets != null) {
return false;
}
} else if (!this.targets.equals(other.targets)) {
return false;
}
if (this.timestampMatrices == null) {
if (other.timestampMatrices != null) {
return false;
}
} else if (!this.timestampMatrices.equals(other.timestampMatrices)) {
return false;
}
if (this.valueMatrices == null) {
if (other.valueMatrices != null) {
return false;
}
} else if (!this.valueMatrices.equals(other.valueMatrices)) {
return false;
}
return true;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset/TimeSeriesDataset2.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset;
import java.util.List;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.util.ClassMapper;
/**
* Dataset for time series.
*
* <p>
* The dataset consists of a value matrices and timestamp matrices. In the
* univariate case, there exists one value matrix, either with or without a
* corresponding timestamp matrix. In the multivariate case there exists
* multiple value matrices, each either with or without a corresponding
* timestamp matrix. Each value matrix is associated with an integer index. Each
* timestamp matrix is associated with an integer index. Two corresponding value
* and timestamp matrices are associated with the same index. The dimensions of
* two corresponding value and timestamp matrices are assured to be the same.
* All value matrices have the same number of rows, but not necessarily the same
* number of columns. The <code>i</code>-th row of each matrix corresponds to
* the <code>i</code>-th instance of the dataset.
* <p>
* <p>
* The targets contained in this dataset are always integers. The can be mapped
* back and forth with the {@link ClassMapper}. Targets are represented as an
* integer array. The <code>i</code>-th entry of this array corresponds to the
* <code>i</code>-th instance of the dataset.
* </p>
*
* @author fischor
*/
public class TimeSeriesDataset2 {
/** Number of instances contained in the dataset. */
private int numberOfInstances;
/** Values of time series variables. */
private List<double[][]> valueMatrices;
/** Timestamps of time series variables. */
private List<double[][]> timestampMatrices;
/** Target values for the instances. */
private int[] targets;
/** States, whether in train (or test) mode. */
private final boolean train;
/**
* Creates a time series dataset with timestamps for training. Let `n` be the
* number of instances.
*
* @param valueMatrices Values for the time series variables. List of
* 2D-Arrays with shape `[n, ?]`.
* @param timestampMatrices Timestamps for the time series variables. List of
* 2D-Arrays with shape `[n, ?]`. Or `null` if no
* timestamps exist for the corresponding time series
* variable. The shape of the `i`th index must be equal
* to the shape of the `i`th element of
* `valueMatrices`.
* @param targets Target values for the instances.
*/
public TimeSeriesDataset2(final List<double[][]> valueMatrices, final List<double[][]> timestampMatrices, final int[] targets) {
// Parameter checks.
// ..
this.numberOfInstances = valueMatrices.get(0).length;
this.valueMatrices = valueMatrices;
this.timestampMatrices = timestampMatrices;
this.targets = targets;
this.train = true;
}
/**
* Creates a time series dataset with timestamps for testing. Let `n` be the
* number of instances.
*
* @param valueMatrices Valueso for the time series variables. List of
* 2D-Arrays with shape `[n, ?]`.
* @param timestampMatrices Timestamps for the time series variables. List of
* 2D-Arrays with shape `[n, ?]`. Or `null` if no
* timestamps exist for the corresponding time series
* variable. The shape of the `i`th index must be equal
* to the shape of the `i`th element of
* `valueMatrices`.
* @param targets Target values for the instances.
*/
public TimeSeriesDataset2(final List<double[][]> valueMatrices, final List<double[][]> timestampMatrices) {
// Parameter checks.
// ..
this.numberOfInstances = valueMatrices.get(0).length;
this.valueMatrices = valueMatrices;
this.timestampMatrices = timestampMatrices;
this.targets = new int[this.numberOfInstances];
this.train = false;
}
/**
* Creates a time series dataset withot timestamps for training. Let `n` be the
* number of instances.
*
* @param valueMatrices Values for the time series variables. List of
* 2D-Arrays with shape `[n, ?]`.
* @param timestampMatrices Timestamps for the time series variables. List of
* 2D-Arrays with shape `[n, ?]`. Or `null` if no
* timestamps exist for the corresponding time series
* variable. The shape of the `i`th index must be equal
* to the shape of the `i`th element of
* `valueMatrices`.
* @param targets Target values for the instances.
*/
public TimeSeriesDataset2(final List<double[][]> valueMatrices, final int[] targets) {
// Parameter checks.
// ..
this(valueMatrices, null, targets);
}
/**
* Creates a time series dataset without timestamps for testing. Let `n` be the
* number of instances.
*
* @param valueMatrices Values for the time series variables. List of 2D-Arrays
* with shape `[n, ?]`.
* @param targets Target values for the instances.
*/
public TimeSeriesDataset2(final List<double[][]> valueMatrices) {
// Parameter checks.
// ..
this.numberOfInstances = valueMatrices.get(0).length;
this.valueMatrices = valueMatrices;
this.timestampMatrices = null;
this.targets = new int[this.numberOfInstances];
this.train = false;
}
/**
* Add a time series variable with timestamps to the dataset.
*
* @param valueMatrix Values for the time series variable to add. 2D-Arrays
* with shape `[n, ?]` where `n` is the number of
* instances of the dataset.
* @param timestampMatrix Timestamps for the time series variable to add.
* 2D-Arrays with shape `[n, ?]` where `n` is the number
* of instances of the dataset. Or `null` if no timestamp
* exists for this time series variable.
*/
public void add(final double[][] valueMatrix, final double[][] timestampMatrix) {
// Parameter checks.
// ..
this.valueMatrices.add(valueMatrix);
this.timestampMatrices.add(timestampMatrix);
}
/**
* Add a time series variable without timestamps to the dataset.
*
* @param valueMatrix Values for the time series variable to add. 2D-Arrays with
* shape `[n, ?]` where `n` is the number of instances of the
* dataset.
*/
public void add(final double[][] valueMatrix) {
// Parameter checks.
// ..
this.valueMatrices.add(valueMatrix);
this.timestampMatrices.add(null);
}
/**
* Removes the time series variable at a given index.
*
* @param index
* @throws IndexOutOfBoundsException
*/
public void remove(final int index) {
this.valueMatrices.remove(index);
this.timestampMatrices.remove(index);
}
/**
* Replaces the time series variable at a given index with a new one.
*
* @param index Index of the time series varialbe to replace.
* @param valueMatrix Values for the time series variable to add. 2D-Arrays
* with shape `[n, ?]` where `n` is the number of
* instances of the dataset.
* @param timestampMatrix Timestamps for the time series variable to add.
* 2D-Arrays with shape `[n, ?]` where `n` is the number
* of instances of the dataset. Or `null` if no timestamp
* exists for this time series variable.
* @throws IndexOutOfBoundsException Thrown if `numberOfInstances <= index`.
*/
public void replace(final int index, final double[][] valueMatrix, final double[][] timestampMatrix) {
this.valueMatrices.set(index, valueMatrix);
if (timestampMatrix != null && this.timestampMatrices != null && this.timestampMatrices.size() > index) {
this.timestampMatrices.set(index, timestampMatrix);
}
}
/**
* Getter for the target values.
*
* @return The targets.
*/
public int[] getTargets() {
return this.targets;
}
/**
* Returns the number of variables, i.e. the number of value matrices contained
* in the dataset.
*
* @return The number of variables.
*/
public int getNumberOfVariables() {
return this.valueMatrices.size();
}
/**
* Returns the number of instances contained in the dataset.
*
* @return The number of instances contained in the dataset.
*/
public int getNumberOfInstances() {
return this.numberOfInstances;
}
/**
* Getter for the value matrix at a specific index. Throws an exception if no
* timestamp matrix exists at this index.
*
* @param index The index of the value matrix.
* @return The value matrix at index <code>index</code>.
* @throws IndexOutOfBoundsException If there is no value matrix at index
* <code>index</code>.
*/
public double[][] getValues(final int index) {
return this.valueMatrices.get(index);
}
/**
* Getter for the timestamp matrix at a specific index.
*
* @param index The index of the timestamp matrix.
* @return The timestamp matrix at index <code>index</code>.
* @throws IndexOutOfBoundsException If there is no value timestamp at index
* <code>index</code>.
*/
public double[][] getTimestamps(final int index) {
return this.timestampMatrices.get(index);
}
/**
* Getter for the value matrix at a specific index.
*
* @param index The index of the timestamp matrix.
* @return The value matrix at index <code>index</code>. Or <code>null</code>,
* if no value matrix exists at index <code>index</code>.
*/
public double[][] getValuesOrNull(final int index) {
return this.valueMatrices.size() > index ? this.valueMatrices.get(index) : null;
}
/**
* Getter for the timestamp matrix at a specific index.
*
* @param index The index of the timestamp matrix.
* @return The timestamp matrix at index <code>index</code>. Or
* <code>null</code>, if no timestamp matrix exists at index
* <code>index</code>.
*/
public double[][] getTimestampsOrNull(final int index) {
if (this.timestampMatrices == null) {
return null;
}
return this.timestampMatrices.size() > index ? this.timestampMatrices.get(index) : null;
}
/**
* States whether the dataset is empty, i.e. contains no value matrices, or not.
*
* @return <code>True</code>, if the dataset is empty. <code>False</code>,
* otherwise.
*/
public boolean isEmpty() {
return this.valueMatrices.isEmpty();
}
/**
* States whether the dataset is a univariate dataset, i.e. contains exactly one
* value matrix, or not.
*
* @return <code>True</code>, if the dataset is univariate. <code>False</code>,
* otherwise.
*/
public boolean isUnivariate() {
return this.valueMatrices.size() == 1;
}
/**
* States whether the dataset is a univariate dataset, i.e. contains more than
* one value matrix, or not.
*
* @return <code>True</code>, if the dataset is multivariate.
* <code>False</code>, otherwise.
*/
public boolean isMultivariate() {
return this.valueMatrices.size() > 1;
}
/**
* States whether the dataset is a training dataset, i.e. contains valid targets
* after initialization, or not.
*
* @return <code>True</code>, if the dataset is a training dataset.
* <code>False</code>, otherwise.
*/
public boolean isTrain() {
return this.train;
}
/**
* States whether the dataset is a test dataset, i.e. contains no valid targets
* after initialization, or not.
*
* @return <code>True</code>, if the dataset is a test dataset.
* <code>False</code>, otherwise.
*/
public boolean isTest() {
return !this.train;
}
/**
* Getter for {@link TimeSeriesDataset2#valueMatrices}.
*
* @return the valueMatrices
*/
public List<double[][]> getValueMatrices() {
return this.valueMatrices;
}
/**
* Getter for {@link TimeSeriesDataset2#timestampMatrices}.
*
* @return the timestampMatrices
*/
public List<double[][]> getTimestampMatrices() {
return this.timestampMatrices;
}
/**
* Setter for {@link TimeSeriesDataset2#valueMatrices}.
*
* @param valueMatrices the valueMatrices to set
*/
public void setValueMatrices(final List<double[][]> valueMatrices) {
this.valueMatrices = valueMatrices;
}
/**
* Setter for {@link TimeSeriesDataset2#timestampMatrices}.
*
* @param timestampMatrices the timestampMatrices to set
*/
public void setTimestampMatrices(final List<double[][]> timestampMatrices) {
this.timestampMatrices = timestampMatrices;
}
/**
* Setter for {@link TimeSeriesDataset2#targets}.
*
* @param targets the targets to set
*/
public void setTargets(final int[] targets) {
this.targets = targets;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset/TimeSeriesFeature.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.util.MathUtil;
/**
* Class calculating features (e. g. mean, stddev or slope) on given
* subsequences of time series. Used e. g. for {@link TimeSeriesTreeClassifier}
* classifier.
*
* @author Julian Lienen
*
*/
public class TimeSeriesFeature {
/**
* Feature types used within the time series tree.
*/
public enum FeatureType {
MEAN, STDDEV, SLOPE
}
/**
* Number of features used within the time series tree.
*/
public static final int NUM_FEATURE_TYPES = FeatureType.values().length;
/**
* Function calculating all features occurring in {@link FeatureType} at once
* using an online calculation approach for mean, standard deviation and the
* slope.
*
* @param vector
* The instance's vector which is used to calculate the features
* @param t1
* Start of the interval
* @param t2
* End of the interval (inclusive)
* @param useBiasCorrection
* Indicator whether the bias (Bessel's) correction should be used
* for the standard deviation calculation
* @return Returns an double array of the size
* {@link TimeSeriesTreeLearningAlgorithm#NUM_FEATURE_TYPES} storing the
* generated feature values.
*/
public static double[] getFeatures(final double[] vector, final int t1, final int t2, final boolean useBiasCorrection) {
double[] result = new double[NUM_FEATURE_TYPES];
if (t1 >= vector.length || t2 >= vector.length) {
throw new IllegalArgumentException("Parameters t1 and t2 must be valid indices of the vector.");
}
if (t1 == t2) {
return new double[] { vector[t1], 0d, 0d };
}
// Calculate running sums for mean, stddev and slope
double xx = 0;
double x = 0;
double xy = 0;
double y = 0;
double yy = 0;
for (int i = t1; i <= t2; i++) {
x += i;
y += vector[i];
yy += vector[i] * vector[i];
xx += i * i;
xy += i * vector[i];
}
double length = t2 - t1 + 1d;
// Calculate the mean
result[0] = y / length;
// Calculate the standard deviation
double stddev = (yy / length - ((y / length) * (y / length)));
if (useBiasCorrection) {
stddev *= length / (length - 1);
}
stddev = Math.sqrt(stddev);
result[1] = stddev;
// Calculate slope
double divisor = (length * xx - x * x);
if (divisor == 0) {
throw new IllegalStateException("Divisor is 0, which must not happen!");
}
result[2] = (length * xy - x * y) / divisor;
return result;
}
/**
* Function calculating the feature specified by the feature type
* <code>fType</code> for a given instance <code>vector</code> of the interval
* [<code>t1</code>, <code>t2</code>].
*
* @param fType
* The feature type to be calculated
* @param instance
* The instance's vector which values are used
* @param t1
* Start of the interval
* @param t2
* End of the interval (inclusive)
* @param useBiasCorrection
* Indicator whether the bias (Bessel's) correction should be used
* for the standard deviation calculation
* @return Returns the calculated feature for the specific instance and interval
*/
public static double calculateFeature(final FeatureType fType, final double[] vector, final int t1, final int t2, final boolean useBiasCorrection) {
switch (fType) {
case MEAN:
return MathUtil.mean(vector, t1, t2);
case STDDEV:
return MathUtil.stddev(vector, t1, t2, useBiasCorrection);
case SLOPE:
return MathUtil.slope(vector, t1, t2);
default:
throw new UnsupportedOperationException("Feature calculation function with id '" + fType + "' is unknwon.");
}
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset/TimeSeriesInstance.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset;
import java.util.Arrays;
import java.util.Iterator;
import java.util.List;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.model.INDArrayTimeseries;
/**
* TimeSeriesInstance
*/
public class TimeSeriesInstance implements ITimeSeriesInstance {
/** Attribute values of the instance. */
private List<INDArrayTimeseries> attributeValues;
/** Target value of the instance. */
private Object label;
/**
* Constructor.
*
* @param dataset
* @param attributeValues
* @param targetValue
*/
public TimeSeriesInstance(final INDArrayTimeseries[] attributeValues, final Object targetValue) {
this(Arrays.asList(attributeValues), targetValue);
}
public TimeSeriesInstance(final List<INDArrayTimeseries> attributeValues, final Object targetValue) {
this.attributeValues = attributeValues;
this.label = targetValue;
}
@Override
public INDArrayTimeseries getAttributeValue(final int pos) {
return this.attributeValues.get(pos);
}
@Override
public Object getLabel() {
return this.label;
}
@Override
public Iterator<INDArrayTimeseries> iterator() {
return this.attributeValues.iterator();
}
@Override
public double[] getPoint() {
double[] point = new double[this.attributeValues.stream().mapToInt(INDArrayTimeseries::length).sum()];
int i = 0;
for (INDArrayTimeseries series : this.attributeValues) {
double[] seriesPoint = series.getPoint();
for (int j = 0; j < seriesPoint.length; j++) {
point[i++] = seriesPoint[j];
}
}
return point;
}
@Override
public Object[] getAttributes() {
return this.attributeValues.toArray();
}
@Override
public void removeColumn(final int columnPos) {
if (columnPos < this.attributeValues.size() && columnPos >= 0) {
this.attributeValues.remove(columnPos);
} else {
throw new IllegalArgumentException("The index is not valid.");
}
}
@Override
public double getPointValue(final int pos) {
throw new UnsupportedOperationException("This operation is not supported.");
}
@Override
public void setLabel(final Object label) {
this.label = label;
}
@Override
public void setAttributeValue(final int pos, final Object value) {
if (!(value instanceof INDArrayTimeseries)) {
throw new IllegalArgumentException("The given value is no timeseries.");
}
this.attributeValues.add((INDArrayTimeseries) value);
}
@Override
public boolean isLabelPresent() {
return this.label != null;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset/package-info.java
|
/**
*
*/
/**
* @author mwever
*
*/
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset;
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset/attribute/ATimeseriesAttribute.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.attribute;
import org.api4.java.ai.ml.core.dataset.schema.attribute.ITimeseries;
import org.api4.java.ai.ml.core.dataset.schema.attribute.ITimeseriesAttribute;
import ai.libs.jaicore.ml.core.dataset.schema.attribute.AGenericObjectAttribute;
/**
*
* @author mwever
*
* @param <O> Type of the representation of a timeseries.
*/
public abstract class ATimeseriesAttribute<O> extends AGenericObjectAttribute<ITimeseries<O>> implements ITimeseriesAttribute<O> {
/**
*
*/
private static final long serialVersionUID = -3411560349820853762L;
private int length;
protected ATimeseriesAttribute(final String name, final int length) {
super(name);
this.length = length;
}
/**
* Get the length of this time series attribute type.
*
* @return The length respec. the number of datapoints of this time series
* attribute.
*/
public int getLength() {
return this.length;
}
public void setLength(final int length) {
if (length < 0) {
throw new IllegalArgumentException("the length has to be greater than or equal to zero.");
}
this.length = length;
}
@Override
public int hashCode() {
final int prime = 31;
int result = super.hashCode();
result = prime * result + this.length;
return result;
}
@Override
public boolean equals(final Object obj) {
if (this == obj) {
return true;
}
if (!super.equals(obj)) {
return false;
}
if (this.getClass() != obj.getClass()) {
return false;
}
ATimeseriesAttribute other = (ATimeseriesAttribute) obj;
return this.length == other.length;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset/attribute/NDArrayTimeseriesAttribute.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.attribute;
import org.api4.java.ai.ml.core.dataset.schema.attribute.ITimeseries;
import org.api4.java.ai.ml.core.dataset.schema.attribute.ITimeseriesAttributeValue;
import org.api4.java.ai.ml.core.dataset.schema.attribute.NoValidAttributeValueException;
import org.nd4j.linalg.api.ndarray.INDArray;
/**
* Describes a time series type as an 1-NDArray with a fixed length.
*/
public class NDArrayTimeseriesAttribute extends ATimeseriesAttribute<INDArray> {
private static final String MSG_NOTIMPLEMENTED = "Not yet implemented";
/**
*
*/
private static final long serialVersionUID = -9188360800052241944L;
public NDArrayTimeseriesAttribute(final String name, final int length) {
super(name, length);
}
/**
* Validates whether a INDArray conforms to this time series. An INDArray
* confirms to this value, if its rank is 1 and its length equals the length of
* this time series.
*
* @param value The value to validated.
* @return Returns true if the given value conforms
*/
@Override
public boolean isValidValue(final Object value) {
if (value instanceof ITimeseries<?> && ((ITimeseries<?>) value).getValue() instanceof INDArray) {
INDArray castedValue = (INDArray) ((ITimeseries<?>) value).getValue();
return castedValue.rank() == 1 && castedValue.length() == this.getLength();
}
return value instanceof NDArrayTimeseriesAttributeValue;
}
@Override
public String getStringDescriptionOfDomain() {
return "[NDATS] " + this.getName();
}
@Override
public ITimeseriesAttributeValue<INDArray> getAsAttributeValue(final Object object) {
return new NDArrayTimeseriesAttributeValue(this, this.getValueAsTypeInstance(object));
}
@SuppressWarnings("unchecked")
@Override
protected ITimeseries<INDArray> getValueAsTypeInstance(final Object object) {
if (this.isValidValue(object)) {
if (object instanceof NDArrayTimeseriesAttributeValue) {
return ((NDArrayTimeseriesAttributeValue) object).getValue();
} else {
return (ITimeseries<INDArray>) object;
}
} else {
throw new NoValidAttributeValueException();
}
}
@Override
public double toDouble(final Object object) {
throw new UnsupportedOperationException(MSG_NOTIMPLEMENTED);
}
@Override
public String serializeAttributeValue(final Object value) {
throw new UnsupportedOperationException(MSG_NOTIMPLEMENTED);
}
@Override
public Object deserializeAttributeValue(final String string) {
throw new UnsupportedOperationException(MSG_NOTIMPLEMENTED);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset/attribute/NDArrayTimeseriesAttributeValue.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.attribute;
import org.api4.java.ai.ml.core.dataset.schema.attribute.ITimeseries;
import org.api4.java.ai.ml.core.dataset.schema.attribute.ITimeseriesAttribute;
import org.api4.java.ai.ml.core.dataset.schema.attribute.ITimeseriesAttributeValue;
import org.nd4j.linalg.api.ndarray.INDArray;
public class NDArrayTimeseriesAttributeValue implements ITimeseriesAttributeValue<INDArray> {
private ITimeseriesAttribute<?> attribute;
private ITimeseries<INDArray> value;
public NDArrayTimeseriesAttributeValue(final ITimeseriesAttribute<?> attribute, final ITimeseries<INDArray> value) {
this.value = value;
this.attribute = attribute;
}
@Override
public ITimeseries<INDArray> getValue() {
return this.value;
}
@Override
public ITimeseriesAttribute<?> getAttribute() {
return this.attribute;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/dataset/attribute/package-info.java
|
/**
*
*/
/**
* @author mwever
*
*/
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.attribute;
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/exception/NoneFittedFilterExeception.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.exception;
import ai.libs.jaicore.ml.core.exception.UncheckedJaicoreMLException;
public class NoneFittedFilterExeception extends UncheckedJaicoreMLException {
/**
*
*/
private static final long serialVersionUID = 1L;
public NoneFittedFilterExeception(final String message, final Throwable cause) {
super(message, cause);
}
public NoneFittedFilterExeception(final String message) {
super(message);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/exception/TimeSeriesLengthException.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.exception;
/**
* Exception class encapsultes faulty behaviour with length of time series.
*
* @author fischor
*/
public class TimeSeriesLengthException extends IllegalArgumentException {
private static final long serialVersionUID = 1L;
public TimeSeriesLengthException(String message, Throwable cause) {
super(message, cause);
}
public TimeSeriesLengthException(String message) {
super(message);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/exception/TimeSeriesLoadingException.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.exception;
/**
* Exception thrown when a time series dataset could not be extracted from an
* external data source (e. g. a file).
*
* @author Julian Lienen
*
*/
public class TimeSeriesLoadingException extends Exception {
/**
* Default generated serial version UID.
*/
private static final long serialVersionUID = -3825008730451093690L;
/**
* Constructor using a nested <code>Throwable</code> exception.
*
* @param message
* Individual exception message
* @param cause
* Nested exception
*/
public TimeSeriesLoadingException(final String message, final Throwable cause) {
super(message, cause);
}
/**
* Standard constructor.
*
* @param message
* Individual exception message
*/
public TimeSeriesLoadingException(final String message) {
super(message);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/filter/AFilter.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.filter;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
public abstract class AFilter implements IFilter {
@Override
public TimeSeriesDataset2 fitTransform(final TimeSeriesDataset2 input) {
this.fit(input);
return this.transform(input);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/filter/IFilter.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.filter;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.exception.NoneFittedFilterExeception;
public interface IFilter {
/**
* represents a function working on a dataset by transforming the dataset itself.
*
* @param input the data set that is to transform
* @return the transformt dataset
* @throws NoneFittedFilterExeception used if transform is called without fit
* @throws IllegalArgumentException used if dataset to transform is empty
*/
public TimeSeriesDataset2 transform(TimeSeriesDataset2 input);
/**
* This function transforms only a single instance.
*
* @param input the to transform instance
* @return the transformed instance
* @throws NoneFittedFilterExeception
*/
public double[] transform(double[] input);
public double[][] transform(double[][] input);
/**
* the function computes the needed information for the transform function.
*
* @param input the dataset that is to transform
*/
public void fit(TimeSeriesDataset2 input);
/**
* The function only fits a single instance of the dataset
*
* @param input The to fit instance
*/
public void fit(double[] input);
public void fit(double[][] input);
/**
* a utility function to avoid the added effort of calling the fit and transform
* function separate
*
* @param input the dataset that is to be transfromed
* @return the transformed dataset
* @throws NoneFittedFilterExeception used if transform is called without fit
* @throws IllegalArgumentException used if dataset to transform is empty
*/
public TimeSeriesDataset2 fitTransform(TimeSeriesDataset2 input);
/**
* the function fit and transforms a single instance
*
* @param input the to fit and transform instance
* @return the transformed instance
*/
public double[] fitTransform(double[] input);
public double[][] fitTransform(double[][] input);
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/filter/SAX.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.filter;
import java.util.Arrays;
import ai.libs.jaicore.basic.transform.vector.PiecewiseAggregateApproximationTransform;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.exception.NoneFittedFilterExeception;
public class SAX implements IFilter {
private double[] alphabet;
private boolean fitted;
private int wordLength;
private double[][] lookuptable;
private ZTransformer ztransform;
private PiecewiseAggregateApproximationTransform paa;
public SAX(final double[] alphabet, final int wordLength) {
this.ztransform = new ZTransformer();
this.alphabet = alphabet;
this.wordLength = wordLength;
this.paa = new PiecewiseAggregateApproximationTransform(wordLength);
}
@Override
public TimeSeriesDataset2 transform(final TimeSeriesDataset2 input) {
if (!(input instanceof TimeSeriesDataset2)) {
throw new IllegalArgumentException("This method only supports TimeSeriesDatasets");
}
if (input.isEmpty()) {
throw new IllegalArgumentException("This method can not work with an empty dataset.");
}
if (!this.fitted) {
throw new NoneFittedFilterExeception("Fit() must be called before transform()");
}
TimeSeriesDataset2 sAXTransformedDataset = new TimeSeriesDataset2(null, null, null);
for (int matrix = 0; matrix < input.getNumberOfVariables(); matrix++) {
double[][] newMatrix = new double[input.getNumberOfInstances()][this.wordLength];
for (int instance = 0; instance < input.getNumberOfInstances(); instance++) {
double[] ppaOfInstance = this.paa.transform(input.getValues(matrix)[instance]);
double[] tsasString = new double[this.wordLength];
double[] localLookupTable = this.lookuptable[matrix];
for (int i = 0; i < ppaOfInstance.length; i++) {
double ppaValue = ppaOfInstance[i];
boolean valuefound = false;
for (int j = 0; j < localLookupTable.length; j++) {
if (ppaValue < localLookupTable[j]) {
tsasString[i] = this.alphabet[j];
valuefound = true;
}
}
if (!valuefound) {
tsasString[i] = this.alphabet[this.alphabet.length - 1];
}
}
newMatrix[instance] = tsasString;
}
sAXTransformedDataset.add(newMatrix, null);
}
return sAXTransformedDataset;
}
@Override
public void fit(final TimeSeriesDataset2 input) {
if (!(input instanceof TimeSeriesDataset2)) {
throw new IllegalArgumentException("This method only supports Timeseriesdatasets");
}
if (input.isEmpty()) {
throw new IllegalArgumentException("This method can not work with an empty dataset.");
}
double[][] maxAndMin = new double[2][input.getNumberOfVariables()];
this.ztransform.fitTransform(input);
for (int matrix = 0; matrix < input.getNumberOfVariables(); matrix++) {
double[] max = new double[input.getNumberOfInstances()];
double[] min = new double[input.getNumberOfInstances()];
for (int instance = 0; instance < input.getNumberOfInstances(); instance++) {
max[instance] = Arrays.stream(input.getValues(matrix)[instance]).max().getAsDouble();
min[instance] = Arrays.stream(input.getValues(matrix)[instance]).min().getAsDouble();
}
maxAndMin[0][matrix] = Arrays.stream(max).max().getAsDouble();
maxAndMin[1][matrix] = Arrays.stream(min).min().getAsDouble();
}
// filling the lookuptable
this.lookuptable = new double[input.getNumberOfVariables()][this.alphabet.length];
for (int matrix = 0; matrix < input.getNumberOfVariables(); matrix++) {
double[] localMaxMin = new double[] { maxAndMin[0][matrix], maxAndMin[1][matrix] };
double totalsize = localMaxMin[0] - localMaxMin[1];
double stepsize = totalsize / this.alphabet.length;
this.lookuptable[matrix][0] = localMaxMin[1] + stepsize;
for (int i = 1; i < this.alphabet.length; i++) {
this.lookuptable[matrix][i] = this.lookuptable[matrix][i - 1] + stepsize;
}
}
this.fitted = true;
}
@Override
public TimeSeriesDataset2 fitTransform(final TimeSeriesDataset2 input) {
this.fit(input);
return this.transform(input);
}
@Override
public double[] transform(final double[] input) {
throw new UnsupportedOperationException();
}
@Override
public void fit(final double[] input) {
throw new UnsupportedOperationException();
}
@Override
public double[] fitTransform(final double[] input) {
throw new UnsupportedOperationException();
}
@Override
public double[][] transform(final double[][] input) {
throw new UnsupportedOperationException();
}
@Override
public void fit(final double[][] input) {
throw new UnsupportedOperationException();
}
@Override
public double[][] fitTransform(final double[][] input) {
throw new UnsupportedOperationException();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/filter/SFA.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.filter;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.List;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.exception.NoneFittedFilterExeception;
/**
* @author Helen Beierling
* c.f. p. 1511 p. 1510 "The BOSS is concerned with time series classification in the presence of noise" by Patrick Schäfer
* This class combines the MCB of finding the bins for a given set of DFT coefficients and SFA
* which selects the right letter for a DFT coefficient.
*/
public class SFA implements IFilter {
private static final String MSG_NOEMPTYDS = "This method can not work with an empty dataset.";
private static final String MSG_NOSINGLEINSTANCE = "To build a SFA word the full dataset has to be considerd therefore it is not reasonable in this context to perform this operation on a single Instance.";
private double[] alphabet;
private boolean meanCorrected;
private boolean fitted = false;
private boolean fittedMatrix = false;
private TimeSeriesDataset2 dFTDataset = null;
private double[][] dftMatrix = null;
private int numberOfDesieredDFTCoefficients;
private List<double[][]> lookupTable = new ArrayList<>();
private double[][] lookUpTableMatrix = null;
private boolean rekursiv;
public void setNumberOfDesieredDFTCoefficients(final int numberOfDesieredDFTCoefficients) {
this.numberOfDesieredDFTCoefficients = numberOfDesieredDFTCoefficients;
}
public void disableRekursiv() {
this.rekursiv = false;
}
public void enableRekursiv() {
this.rekursiv = true;
}
public SFA(final double[] alphabet, final int wordLength) {
this.alphabet = alphabet;
// The wordlength must be even
this.numberOfDesieredDFTCoefficients = wordLength / 2;
}
@Override
public TimeSeriesDataset2 transform(final TimeSeriesDataset2 input) {
if (input.isEmpty()) {
throw new IllegalArgumentException(MSG_NOEMPTYDS);
}
if (!this.fitted) {
throw new NoneFittedFilterExeception("The filter must be fitted before it can transform.");
}
List<double[][]> sfaDataset = new ArrayList<>();
// calculate SFA words for every instance and its DFT coefficients
for (int matrix = 0; matrix < this.dFTDataset.getNumberOfVariables(); matrix++) {
double[][] sfaWords = new double[this.dFTDataset.getNumberOfInstances()][this.numberOfDesieredDFTCoefficients * 2];
for (int instance = 0; instance < this.dFTDataset.getNumberOfInstances(); instance++) {
for (int entry = 0; entry < this.numberOfDesieredDFTCoefficients * 2; entry++) {
double elem = this.dFTDataset.getValues(matrix)[instance][entry];
// get the lookup table for DFT values of the instance
double[] lookup = this.lookupTable.get(matrix)[entry];
// if the DFT coefficient is smaller than the first or larger than the last
// or it lays on the border give it first, last or second ,penultimate
if (elem < lookup[0]) {
sfaWords[instance][entry] = this.alphabet[0];
} else if (elem == lookup[0]) {
sfaWords[instance][entry] = this.alphabet[1];
} else if (elem > lookup[this.alphabet.length - 2]) {
sfaWords[instance][entry] = this.alphabet[this.alphabet.length - 1];
} else if (elem == lookup[this.alphabet.length - 2]) {
sfaWords[instance][entry] = this.alphabet[this.alphabet.length - 1];
}
// get alphabet letter for every non extreme coefficient
else {
for (int i = 1; i < lookup.length; i++) {
if (elem <= lookup[i]) {
if (elem < lookup[i]) {
sfaWords[instance][entry] = this.alphabet[i];
} else if (elem == lookup[i]) {
sfaWords[instance][entry] = this.alphabet[i + 1];
}
break;
}
}
}
}
}
sfaDataset.add(sfaWords);
}
return new TimeSeriesDataset2(sfaDataset, null, null);
}
@Override
public void fit(final TimeSeriesDataset2 input) {
if (input.isEmpty()) {
throw new IllegalArgumentException(MSG_NOEMPTYDS);
}
if (this.alphabet.length == 0) {
throw new IllegalArgumentException("The alphabet size can not be zero.");
}
this.lookupTable.clear();
DFT dftFilter = new DFT();
dftFilter.setMeanCorrected(this.meanCorrected);
// calculates the number of DFT coefficents with wordlength as number of desired DFT coefficients
dftFilter.setNumberOfDisieredCoefficients(this.numberOfDesieredDFTCoefficients);
if (!this.rekursiv) {
this.dFTDataset = dftFilter.fitTransform(input);
} else {
// Only works for sliding windows. However it is normally used for SFA.
this.dFTDataset = dftFilter.rekursivDFT(input);
}
for (int matrix = 0; matrix < this.dFTDataset.getNumberOfVariables(); matrix++) {
// for each part of every coefficient calculate the bins for the alphabet (number of bins == number of letters)
double[][] lookUpTable = new double[this.numberOfDesieredDFTCoefficients * 2][this.alphabet.length - 1];
for (int coeficient = 0; coeficient < this.numberOfDesieredDFTCoefficients * 2; coeficient++) {
// get the columns of the DFT dataset
double[] toBin = new double[input.getNumberOfInstances()];
for (int instances = 0; instances < this.dFTDataset.getNumberOfInstances(); instances++) {
toBin[instances] = this.dFTDataset.getValues(matrix)[instances][coeficient];
}
// Sort ascending
// If the number of instances is equal to the number of bins the breakpoints are set to this values
Arrays.sort(toBin);
if (toBin.length == this.alphabet.length - 1) {
lookUpTable[coeficient] = Arrays.copyOf(toBin, this.alphabet.length - 1);
}
// If the number of instances is greater than the number of bins then the breakpoints are set
// in the way that all coefficients are spread equally over the bins
else {
int splitValue = (int) Math.round(toBin.length / (double) this.alphabet.length);
for (int alphabetLetter = 1; alphabetLetter < this.alphabet.length; alphabetLetter++) {
lookUpTable[coeficient][alphabetLetter - 1] = toBin[alphabetLetter * splitValue];
}
}
}
this.lookupTable.add(lookUpTable);
}
this.fitted = true;
}
@Override
public TimeSeriesDataset2 fitTransform(final TimeSeriesDataset2 input) {
this.fit(input);
return this.transform(input);
}
@Override
public double[] transform(final double[] input) {
throw new UnsupportedOperationException(MSG_NOSINGLEINSTANCE);
}
@Override
public void fit(final double[] input) {
throw new UnsupportedOperationException(MSG_NOSINGLEINSTANCE);
}
@Override
public double[] fitTransform(final double[] input) {
throw new UnsupportedOperationException(MSG_NOSINGLEINSTANCE);
}
@Override
public double[][] transform(final double[][] input) {
if (input.length == 0) {
throw new IllegalArgumentException(MSG_NOEMPTYDS);
}
if (!this.fittedMatrix) {
throw new NoneFittedFilterExeception("The filter must be fitted before it can transform.");
}
double[][] sfaMatrix = new double[this.dftMatrix.length][this.numberOfDesieredDFTCoefficients * 2];
for (int instance = 0; instance < this.dftMatrix.length; instance++) {
for (int entry = 0; entry < this.numberOfDesieredDFTCoefficients * 2; entry++) {
double elem = this.dftMatrix[instance][entry];
// get the lookup table for DFT values of the instance
double[] lookup = this.lookUpTableMatrix[entry];
// if the DFT coefficient is smaller than the first or larger than the last
// or it lays on the border give it first, last or second ,penultimate
if (elem < lookup[0]) {
sfaMatrix[instance][entry] = this.alphabet[0];
}
if (elem == lookup[0]) {
sfaMatrix[instance][entry] = this.alphabet[1];
}
if (elem > lookup[this.alphabet.length - 2]) {
sfaMatrix[instance][entry] = this.alphabet[this.alphabet.length - 1];
}
if (elem == lookup[this.alphabet.length - 2]) {
sfaMatrix[instance][entry] = this.alphabet[this.alphabet.length - 1];
}
// get alphabet letter for every non extrem coefficient
else {
for (int i = 1; i < lookup.length - 2; i++) {
if (elem > lookup[i]) {
sfaMatrix[instance][entry] = this.alphabet[i];
}
if (elem == lookup[i]) {
sfaMatrix[instance][entry] = this.alphabet[i + 1];
}
}
}
}
}
return sfaMatrix;
}
/*
* Can not be called in the fit dataset method because it needs its own DFT-
* Filter and for the dataset an overall Filter is needed.
*/
@Override
public void fit(final double[][] input) {
if (input.length == 0) {
throw new IllegalArgumentException(MSG_NOEMPTYDS);
}
if (this.alphabet.length == 0) {
throw new IllegalArgumentException("The alphabet size can not be zero.");
}
DFT dftFilterMatrix = new DFT();
// calculates the number of DFT coefficents with wordlength as number of desired DFT coefficients
dftFilterMatrix.setNumberOfDisieredCoefficients(this.numberOfDesieredDFTCoefficients);
if (!this.rekursiv) {
this.dftMatrix = dftFilterMatrix.fitTransform(input);
} else {
this.dftMatrix = dftFilterMatrix.rekursivDFT(input);
}
this.lookUpTableMatrix = new double[this.numberOfDesieredDFTCoefficients * 2][this.alphabet.length - 1];
for (int coeficient = 0; coeficient < this.numberOfDesieredDFTCoefficients * 2; coeficient++) {
// get the columns of the DFT dataset
double[] toBin = new double[input.length];
for (int instances = 0; instances < input.length; instances++) {
toBin[instances] = this.dftMatrix[instances][coeficient];
}
// Sort ascending
// If the number of instances is equal to the number of bins the breakpoints are set to this values
Arrays.sort(toBin);
if (toBin.length == this.alphabet.length - 1) {
this.lookUpTableMatrix[coeficient] = Arrays.copyOf(toBin, this.alphabet.length - 1);
}
// If the number of instances is greater than the number of bins then the breakpoints are set
// in the way that all coefficients are spread equally over the bins
else {
int splitValue = (int) Math.round(toBin.length / (double) this.alphabet.length);
for (int alphabetLetter = 0; alphabetLetter < this.alphabet.length - 1; alphabetLetter++) {
this.lookUpTableMatrix[coeficient][alphabetLetter] = toBin[alphabetLetter + splitValue];
}
}
}
this.fittedMatrix = true;
}
@Override
public double[][] fitTransform(final double[][] input) {
this.fit(input);
return this.transform(input);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/filter/SlidingWindowBuilder.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.filter;
import java.util.ArrayList;
import java.util.Arrays;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.exception.NoneFittedFilterExeception;
/**
* @author Helen Beierling
* This class cuts an instance or a set of instances into a number of smaller instances which are
* typically saved in an matrix per instance and the matrices in a list.
* c.f. p.1508 "The BOSS is concerned with time series classification in the presence of noise" by Patrick Schaefer
*/
public class SlidingWindowBuilder implements IFilter {
private static final String MSG_SPECIALFIT = "This is done by the special fit and transform because this mehtod must return a new dataset not a double array.";
private boolean fitted = false;
private boolean fittedMatrix = false;
private int defaultWindowSize = 20;
private ArrayList<double[][]> blownUpDataset = new ArrayList<>();
private double[][] blownUpMatrix = null;
public void setDefaultWindowSize(final int defaultWindowSize) {
this.defaultWindowSize = defaultWindowSize;
}
public int getDefaultWindowSize() {
return this.defaultWindowSize;
}
@Override
public TimeSeriesDataset2 transform(final TimeSeriesDataset2 input) {
if (input.isEmpty()) {
throw new IllegalArgumentException("The input dataset can not be empty");
}
if (!this.fitted) {
throw new NoneFittedFilterExeception("The fit method must be called before transformning");
}
return new TimeSeriesDataset2(this.blownUpDataset, null, null);
}
@Override
// Results in a list of matrices where each instance has its own matrix.
// Therefore the structure of the matrices are lost if this method is used.
public void fit(final TimeSeriesDataset2 input) {
for (int matrix = 0; matrix < input.getNumberOfVariables(); matrix++) {
ArrayList<double[][]> newMatrices = new ArrayList<>();
for (double[] instance : input.getValues(matrix)) {
double[][] newMatrix = new double[(instance.length - this.defaultWindowSize)][this.defaultWindowSize];
for (int entry = 0; entry < instance.length - this.defaultWindowSize; entry++) {
double[] tmp = Arrays.copyOfRange(instance, entry, entry + this.defaultWindowSize);
newMatrix[entry] = tmp;
}
newMatrices.add(newMatrix);
}
this.blownUpDataset = newMatrices;
}
this.fitted = true;
}
/**
* This is an extra fit method because it does not return a double[] array even though it gets
* a double [] as input as it would be defined in the .
*
* @param instance that has to be transformed
* @return the tsdataset that results from one instance which consists of
* one matrix with each row represents one part of the instance from i to i+ window length for i < n- window length
*/
public TimeSeriesDataset2 specialFitTransform(final double[] instance) {
if (instance.length == 0) {
throw new IllegalArgumentException("The input instance can not be empty");
}
if (instance.length < this.defaultWindowSize) {
throw new IllegalArgumentException("The input instance can not be smaller than the windowsize");
}
double[][] newMatrix = new double[instance.length - this.defaultWindowSize + 1][this.defaultWindowSize];
for (int entry = 0; entry <= instance.length - (this.defaultWindowSize); entry++) {
newMatrix[entry] = Arrays.copyOfRange(instance, entry, entry + this.defaultWindowSize);
}
ArrayList<double[][]> newDataset = new ArrayList<>();
newDataset.add(newMatrix);
return new TimeSeriesDataset2(newDataset);
}
@Override
public TimeSeriesDataset2 fitTransform(final TimeSeriesDataset2 input) {
this.fit(input);
return this.transform(input);
}
/*
* This operation is unsupported because it would result in one stream of new instances in one array.
*/
@Override
public double[] transform(final double[] input) {
throw new UnsupportedOperationException(MSG_SPECIALFIT);
}
/*
* This method is unsupported because the corresponding transform operation is
* not useful
*/
@Override
public void fit(final double[] input) {
throw new UnsupportedOperationException(MSG_SPECIALFIT);
}
@Override
public double[] fitTransform(final double[] input) {
throw new UnsupportedOperationException(MSG_SPECIALFIT);
}
@Override
public double[][] transform(final double[][] input) {
if (input.length == 0) {
throw new IllegalArgumentException("The input matrix can not be empty");
}
if (!this.fittedMatrix) {
throw new NoneFittedFilterExeception("The fit mehtod must be called before transformning");
}
return this.blownUpMatrix;
}
@Override
// Does not return a list of matrices but a bigger matrix where the new created instances are getting stacked
// if there is a instance of size n than the first n-window length rows are the sliced instance.
public void fit(final double[][] input) {
if (input.length == 0) {
throw new IllegalArgumentException("The input matrix can not be empty");
}
// This is the buffer for the new matrix that gets created from a single instance.
this.blownUpMatrix = new double[input.length * (input[0].length - this.defaultWindowSize)][this.defaultWindowSize];
for (int instance = 0; instance < input.length; instance++) {
for (int entry = 0; entry < input[instance].length - this.defaultWindowSize; entry++) {
// Every entry in the new matrix is equal to a copy of the original instance from
// entry i to entry i plus window length.
this.blownUpMatrix[instance + (entry)] = Arrays.copyOfRange(input[instance], entry, entry + this.defaultWindowSize);
}
}
this.fittedMatrix = true;
}
@Override
public double[][] fitTransform(final double[][] input) {
this.fit(input);
return this.transform(input);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/filter/ZTransformer.java
|
/**
*
*/
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.filter;
import java.util.ArrayList;
import java.util.List;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.exception.NoneFittedFilterExeception;
/**
* @author Helen Beierling
* This class normalizes the mean of an instance to be zero and the deviation to be one.
* s.https://jmotif.github.io/sax-vsm_site/morea/algorithm/znorm.html
* one loop: https://www.strchr.com/standard_deviation_in_one_pass?allcomments=1
*
* XXX: Duplicates functionality in ai.libs.jaicore.basic.transform.vector.ZTransform
*/
public class ZTransformer extends AFilter {
private double mean;
private double deviation;
private List<double[][]> ztransformedDataset = new ArrayList<>();
// To get a unbiased estimate for the variance the intermediated results are
// divided by n-1 instead of n(Number of samples of Population)
private boolean basselCorrected = true;
private boolean fitted = false;
private boolean fittedInstance = false;
private boolean fittedMatrix = false;
public void setBasselCorrected(final boolean basselCorrected) {
this.basselCorrected = basselCorrected;
}
/* (non-Javadoc)
* @see jaicore.ml.tsc.filter.IFilter#transform(jaicore.ml.core.dataset.IDataset)
*/
@Override
public TimeSeriesDataset2 transform(final TimeSeriesDataset2 input) {
if (input.isEmpty()) {
throw new IllegalArgumentException("This method can not work with an empty dataset.");
}
if (!this.fitted) {
throw new NoneFittedFilterExeception("The fit method must be called before the transform method.");
}
for (int matrix = 0; matrix < input.getNumberOfVariables(); matrix++) {
this.ztransformedDataset.add(this.fitTransform(input.getValues(matrix)));
this.fittedMatrix = false;
}
this.fitted = false;
return new TimeSeriesDataset2(this.ztransformedDataset);
}
@Override
public void fit(final TimeSeriesDataset2 input) {
if (input.isEmpty()) {
throw new IllegalArgumentException("This method can not work with an empty dataset.");
}
this.fitted = true;
}
@Override
public TimeSeriesDataset2 fitTransform(final TimeSeriesDataset2 input) {
this.fit(input);
return this.transform(input);
}
@Override
public double[] transform(final double[] input) {
if (!this.fittedInstance) {
throw new NoneFittedFilterExeception("The fit method must be called before the transfom method is called");
}
if (input.length == 0) {
throw new IllegalArgumentException("The to transform instance can not be empty");
}
double[] ztransform = new double[input.length];
for (int entry = 0; entry < input.length; entry++) {
if (this.deviation != 0) {
ztransform[entry] = (entry - this.mean) / this.deviation;
}
}
this.fittedInstance = false;
return ztransform;
}
@Override
public void fit(final double[] input) {
double sumSq = 0.0;
double sumMean = 0.0;
double numberOfEntrys = input.length;
if (numberOfEntrys == 0) {
throw new IllegalArgumentException("The to transform instance can not be empty.");
}
for (int entry = 0; entry < input.length; entry++) {
sumSq = sumSq + Math.pow(input[entry], 2);
sumMean = sumMean + input[entry];
}
this.mean = sumMean / numberOfEntrys;
double variance = (1 / numberOfEntrys) * (sumSq) - Math.pow(this.mean, 2);
if (this.basselCorrected) {
double tmp = (numberOfEntrys / (numberOfEntrys - 1));
this.deviation = Math.sqrt(tmp * variance);
} else {
this.deviation = Math.sqrt(variance);
}
this.fittedInstance = true;
}
@Override
public double[] fitTransform(final double[] input) {
this.fit(input);
return this.transform(input);
}
@Override
public double[][] transform(final double[][] input) {
if (!this.fittedMatrix) {
throw new NoneFittedFilterExeception("The fit method must be called first.");
}
double[][] ztransformedMatrix = new double[input.length][input[0].length];
for (int instance = 0; instance < input.length; instance++) {
ztransformedMatrix[instance] = this.fitTransform(input[instance]);
this.fittedInstance = false;
}
this.fittedMatrix = false;
return ztransformedMatrix;
}
@Override
public void fit(final double[][] input) {
this.fittedMatrix = true;
}
@Override
public double[][] fitTransform(final double[][] input) {
this.fit(input);
return this.transform(input);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner/ASimplifiedTSCLearningAlgorithm.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner;
import java.util.Iterator;
import java.util.NoSuchElementException;
import org.api4.java.algorithm.events.IAlgorithmEvent;
import ai.libs.jaicore.basic.IOwnerBasedAlgorithmConfig;
import ai.libs.jaicore.basic.algorithm.AAlgorithm;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
public abstract class ASimplifiedTSCLearningAlgorithm<T, C extends ASimplifiedTSClassifier<T>> extends AAlgorithm<TimeSeriesDataset2, C> {
protected ASimplifiedTSCLearningAlgorithm(final IOwnerBasedAlgorithmConfig config, final C classifier, final TimeSeriesDataset2 input) {
super(config, input);
this.classifier = classifier; // this is the classifier that is being trained (and outputted in the end)
}
/**
* The model which is maintained during algorithm calls
*/
private final C classifier;
public C getClassifier() {
return this.classifier;
}
/**
* {@inheritDoc}
*/
@Override
public void registerListener(final Object listener) {
throw new UnsupportedOperationException();
}
/**
* {@inheritDoc}
*/
@Override
public IAlgorithmEvent nextWithException() {
throw new UnsupportedOperationException();
}
/**
* {@inheritDoc}
*/
@Override
public Iterator<IAlgorithmEvent> iterator() {
throw new UnsupportedOperationException();
}
/**
* {@inheritDoc}
*/
@Override
public boolean hasNext() {
throw new UnsupportedOperationException();
}
/**
* {@inheritDoc}
*/
@Override
public IAlgorithmEvent next() {
throw new NoSuchElementException("Cannot enumerate on this algorithm");
}
/**
* {@inheritDoc}
*/
@Override
public void cancel() {
throw new UnsupportedOperationException();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner/ASimplifiedTSClassifier.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner;
import java.util.List;
import org.api4.java.ai.ml.core.exception.PredictionException;
import org.api4.java.ai.ml.core.exception.TrainingException;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.util.ClassMapper;
/**
* Simplified batch-learning time series classifier which can be trained and
* used as a predictor. Uses <code>algorithm</code> to train the model
* parameters (if necessary).
*
* @author Julian Lienen
*
*/
public abstract class ASimplifiedTSClassifier<T> {
/**
* Class mapper object used to encode and decode predicted values if String
* values are used as classes. Can be null if the predicted values are not
* mapped to String values.
*/
protected ClassMapper classMapper;
/**
* Variable indicating whether the classifier has been trained.
*/
protected boolean trained;
public ASimplifiedTSClassifier() {
}
/**
* Performs a prediction based on the given univariate double[] instance
* representation and returns the result.
*
* @param univInstance Univariate instance given by a double vector of time
* series values used for the prediction
* @return Returns the result of the prediction
* @throws PredictionException If something fails during the prediction process.
*/
public abstract T predict(final double[] univInstance) throws PredictionException;
/**
* Performs a prediction based on the given multivariate list of double[]
* instance representation and returns the result.
*
* @param multivInstance Multivariate instance given by a list of multiple
* double[] time series used for the prediction
* @return Returns the result of the prediction
* @throws PredictionException If something fails during the prediction process.
*/
public T predict(final List<double[]> multivInstance) throws PredictionException {
throw new PredictionException("Can't predict on multivariate data yet.");
}
/**
* Performs predictions based on the given instances in the given dataset.
*
* @param dataset The {@link TimeSeriesDataset2} for which predictions should be
* made.
* @return Returns the result of the predictions
* @throws PredictionException If something fails during the prediction process
*/
public abstract List<T> predict(final TimeSeriesDataset2 dataset) throws PredictionException;
public abstract <U extends ASimplifiedTSClassifier<T>> ASimplifiedTSCLearningAlgorithm<T, U> getLearningAlgorithm(final TimeSeriesDataset2 dataset);
/**
* Trains the simplified time series classifier model using the given
* {@link TimeSeriesDataset2}.
*
* @param dataset The {@link TimeSeriesDataset2} which should be used for the
* training.
* @throws TrainingException If something fails during the training process.
*/
public final void train(final TimeSeriesDataset2 dataset) throws InterruptedException, TrainingException {
// Set model which is trained
ASimplifiedTSCLearningAlgorithm<T, ? extends ASimplifiedTSClassifier<T>> algorithm = this.getLearningAlgorithm(dataset);
// Set input data from which the model should learn
try {
// Train
algorithm.call();
this.trained = true;
} catch (InterruptedException e) {
throw e;
} catch (Exception e) {
throw new TrainingException("Could not train model " + this.getClass().getSimpleName(), e);
}
}
/**
* Getter for the property <code>classMapper</code>.
*
* @return Returns the actual class mapper or null if no mapper is stored
*/
public ClassMapper getClassMapper() {
return this.classMapper;
}
/**
* Setter for the property <code>classMapper</code>.
*
* @param classMapper The class mapper to be set
*/
public void setClassMapper(final ClassMapper classMapper) {
this.classMapper = classMapper;
}
/**
* @return the trained
*/
public boolean isTrained() {
return this.trained;
}
protected double[][] checkWhetherPredictionIsPossible(final TimeSeriesDataset2 dataset) throws PredictionException {
// Parameter checks.
if (!this.isTrained()) {
throw new PredictionException("Model has not been built before!");
}
if (dataset == null || dataset.isEmpty()) {
throw new IllegalArgumentException("Dataset to be predicted must not be null or empty!");
}
double[][] testInstances = dataset.getValuesOrNull(0);
if (testInstances == null) {
throw new PredictionException("Can't predict on empty dataset.");
}
return testInstances;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner/ATSCAlgorithm.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner;
import org.api4.java.algorithm.IAlgorithm;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset;
/**
* Abstract algorithm class which is able to take {@link TimeSeriesDataset}
* objects and builds {@link ATimeSeriesClassificationModel} instances specified by the generic
* parameter <CLASSIFIER>.
*
* @author Julian Lienen
*
* @param <Y>
* The type of the target that the <CLASSIFIER> to be trained
* @param <V>
* The value type of the target that the <CLASSIFIER> to be trained
* predicts.
* @param <D>
* The type of the time series data set used to learn from and
* predict batches.
* @param <C>
* The time series classifier which is modified and returned as
* algorithm result.
*/
public abstract class ATSCAlgorithm<Y, D extends TimeSeriesDataset, C extends ATimeSeriesClassificationModel<Y, D>> implements IAlgorithm<TimeSeriesDataset, C> {
/**
* The model which is maintained during algorithm calls
*/
protected C model;
/**
* The {@link TimeSeriesDataset} object used for maintaining the
* <code>model</code>.
*/
protected D input;
/**
* Setter for the model to be maintained.
*
* @param model
* The {@link ATimeSeriesClassificationModel} model which is maintained during
* algorithm calls.
*/
@SuppressWarnings("unchecked")
public <T extends ATimeSeriesClassificationModel<Y, D>> void setModel(final T model) {
this.model = (C) model;
}
/**
* Setter for the data set input used during algorithm calls.
*
* @param input
* The {@link TimeSeriesDataset} object (or a subtype) used for the
* model maintenance
*/
public void setInput(final D input) {
this.input = input;
}
/**
* Getter for the data set input used during algorithm calls.
*/
@Override
public D getInput() {
return this.input;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner/ATimeSeriesClassificationModel.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassification;
import org.api4.java.ai.ml.classification.singlelabel.evaluation.ISingleLabelClassificationPredictionBatch;
import org.api4.java.ai.ml.core.exception.TrainingException;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.ITimeSeriesInstance;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset;
import ai.libs.jaicore.ml.core.learner.ASupervisedLearner;
/**
* Time series classifier which can be trained and used as a predictor. Uses
* <code>algorithm</code> to train the model parameters (if necessary).
*
* @author Julian Lienen
*
* @param <L>
* The attribute type of the target.
* @param <V>
* The value type of the target attribute.
* @param <D>
* The type of the time series data set used to learn from and
* predict batches.
*/
public abstract class ATimeSeriesClassificationModel<L, D extends TimeSeriesDataset> extends ASupervisedLearner<ITimeSeriesInstance, D, ISingleLabelClassification, ISingleLabelClassificationPredictionBatch> {
/**
* The algorithm object used for the training of the classifier.
*/
protected ATSCAlgorithm<L, D, ? extends ATimeSeriesClassificationModel<L, D>> algorithm;
/**
* Constructor for a time series classifier.
*
* @param algorithm
* The algorithm object used for the training of the classifier
*/
public ATimeSeriesClassificationModel(final ATSCAlgorithm<L, D, ? extends ATimeSeriesClassificationModel<L, D>> algorithm) {
this.algorithm = algorithm;
}
/**
* {@inheritDoc ABatchLearner#train(jaicore.ml.core.dataset.IDataset)}
*/
@Override
public void fit(final D dataset) throws InterruptedException, TrainingException {
// Set model which is trained
this.algorithm.setModel(this);
// Set input data from which the model should learn
this.algorithm.setInput(dataset);
try {
// Train
this.algorithm.call();
} catch (InterruptedException e) {
throw e;
} catch (Exception e) {
throw new TrainingException("Could not train model " + this.getClass().getSimpleName(), e);
}
}
/**
* Getter for the model's training algorithm object.
*
* @return The model's training algorithm
*/
public ATSCAlgorithm<L, D, ? extends ATimeSeriesClassificationModel<L, D>> getAlgorithm() {
return this.algorithm;
}
/**
* Sets the training algorithm for the classifier.
*
* @param algorithm
* The algorithm object used to maintain the model's parameters.
*/
public void setAlgorithm(final ATSCAlgorithm<L, D, ? extends ATimeSeriesClassificationModel<L, D>> algorithm) {
this.algorithm = algorithm;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner/BOSSClassifier.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner;
import java.util.ArrayList;
import java.util.List;
import java.util.Map;
import java.util.Map.Entry;
import org.aeonbits.owner.ConfigCache;
import org.api4.java.ai.ml.core.exception.PredictionException;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.filter.SFA;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.filter.SlidingWindowBuilder;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.filter.ZTransformer;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner.BOSSLearningAlgorithm.IBossAlgorithmConfig;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.util.HistogramBuilder;
/**
* @author Helen Beierling
* This class predicts labels for instances by cutting the instance into pieces and calculate for every piece the DFT values and
* assigns them to the corresponding letter to form the SFA words of that piece and afterwards creating a histogram for which the
* distance is calculated to all histograms of the training data and the label of the nearest one is returned as prediction.
*/
public class BOSSClassifier extends ASimplifiedTSClassifier<Integer> {
private TimeSeriesDataset2 trainingData;
// ---------------------------------------------------------------
// All variables needed for the to predict instance and for the BOSS Algorithm or calculated by it .
private final IBossAlgorithmConfig config;
private List<Map<Integer, Integer>> univirateHistograms;
// ---------------------------------------------------------------
// All needed for every predict.
private SlidingWindowBuilder slide = new SlidingWindowBuilder();
private HistogramBuilder histoBuilder = new HistogramBuilder();
private ZTransformer znorm = new ZTransformer();
public BOSSClassifier(final int windowLength, final int alphabetSize, final double[] alphabet, final int wordLength, final boolean meanCorrected) {
this.config = ConfigCache.getOrCreate(IBossAlgorithmConfig.class);
this.config.setProperty(IBossAlgorithmConfig.K_WINDOW_SIZE, "" + windowLength);
this.config.setProperty(IBossAlgorithmConfig.K_ALPHABET_SIZE, "" + alphabetSize);
this.config.setProperty(IBossAlgorithmConfig.K_ALPHABET, "" + alphabet);
this.config.setProperty(IBossAlgorithmConfig.K_WORDLENGTH, "" + wordLength);
this.config.setProperty(IBossAlgorithmConfig.K_MEANCORRECTED, "" + meanCorrected);
this.slide.setDefaultWindowSize(this.config.windowSize());
}
public BOSSClassifier(final IBossAlgorithmConfig config) {
this.config = config;
// This is the same window size as used for the training samples
this.slide.setDefaultWindowSize(config.windowSize());
}
public List<Map<Integer, Integer>> getUnivirateHistograms() {
return this.univirateHistograms;
}
public void setTrainingData(final TimeSeriesDataset2 trainingData) {
this.trainingData = trainingData;
}
public void setHistogramUnivirate(final List<Map<Integer, Integer>> histograms) {
this.univirateHistograms = histograms;
}
@Override
public Integer predict(final double[] univInstance) throws PredictionException {
SFA sfa = new SFA(this.config.alphabet(), this.config.wordLength());
// create windows for test instance an there for a small dataset with
// windows as instances.
TimeSeriesDataset2 tmp = this.slide.specialFitTransform(univInstance);
// need to call a new fit for each predict because each window gets z normalized by its own.
// c.f.p. 1509 "The BOSS is concerned with time series classification in the presence of noise by Patrick Schaefer"
for (int instance = 0; instance < tmp.getValues(0).length; instance++) {
tmp.getValues(0)[instance] = this.znorm.fitTransform(tmp.getValues(0)[instance]);
}
TimeSeriesDataset2 tmpznormedsfaTransformed = sfa.fitTransform(tmp);
Map<Integer, Integer> histogram = this.histoBuilder.histogramForInstance(tmpznormedsfaTransformed);
// Calculate distance for all histograms for all instances in the training set.
// Remember index of histogram with minimum distance in list because it corresponds to the
// instance that produced that histogram with minimum distance.
int indexOFminDistInstance = 0;
double minDist = Double.MAX_VALUE;
for (int i = 0; i < this.univirateHistograms.size(); i++) {
double dist = this.getBossDistance(histogram, this.univirateHistograms.get(i));
if (dist < minDist) {
minDist = dist;
indexOFminDistInstance = i;
}
}
// return the target of that instance that had the minimum distance.
return this.trainingData.getTargets()[indexOFminDistInstance];
}
@Override
public Integer predict(final List<double[]> multivInstance) throws PredictionException {
// The BOSS classifier only supports predictions for univariate instances.
throw new UnsupportedOperationException("The BOSS classifier is a univariat classifer");
}
@Override
public List<Integer> predict(final TimeSeriesDataset2 dataset) throws PredictionException {
// For a list of instances a list of predictions are getting created and the list is than returned.
List<Integer> predictions = new ArrayList<>();
for (double[][] matrix : dataset.getValueMatrices()) {
for (double[] instance : matrix) {
predictions.add(this.predict(instance));
}
}
return predictions;
}
/**
* @param a The distance starting point histogram.
* @param b The distance destination histogram.
* @return The distance between Histogram a and b.
*
* The distance itself is calculated as 0 if the word does appear in "b" but not in "a" and
* if the word exists in "a" but not in "b" it is the word count of "a" squared. For the "normal" case
* where the word exists in "a" and "b" the distance is word count of "a" minus "b" and the result gets
* squared for each word and summed up over the whole histogram.
* Therefore the two histograms do not need to be of the same size and the distance of "a" to "b" must not
* be equal to the distance of "b" to "a".
* c.f. p. 1516 "The BOSS is concerned with time series classification in the presence of noise by Patrick Schaefer"
*/
private double getBossDistance(final Map<Integer, Integer> a, final Map<Integer, Integer> b) {
double result = 0;
for (Entry<Integer, Integer> entry : a.entrySet()) {
int key = entry.getKey();
int val = entry.getValue();
if (b.containsKey(key)) {
result += (Math.pow(val - (double) b.get(key), 2));
} else {
result += Math.pow(val, 2);
}
}
return result;
}
@Override
public BOSSLearningAlgorithm getLearningAlgorithm(final TimeSeriesDataset2 dataset) {
return new BOSSLearningAlgorithm(this.config, this, dataset);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner/BOSSEnsembleClassifier.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Map.Entry;
import org.api4.java.ai.ml.core.exception.PredictionException;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
/* This class is just a sketch for the BOSS ensemble classifier it assumes that the grid
* of the parameters window size and word length is already computed
* and that the best ones according to a percentage of the best combination are already chosen
* and put into the delivered HashMap.
* cf.p.1520
* "The BOSS is concerned with time series classification in the presence of noise by Patrick Schaefer" */
public class BOSSEnsembleClassifier extends ASimplifiedTSClassifier<Integer> {
private ArrayList<BOSSClassifier> ensemble = new ArrayList<>();
public BOSSEnsembleClassifier(final Map<Integer, Integer> windowLengthsandWordLength, final int alphabetSize, final double[] alphabet, final boolean meanCorrected) {
for (Entry<Integer, Integer> lengthPair : windowLengthsandWordLength.entrySet()) {
this.ensemble.add(new BOSSClassifier(lengthPair.getKey(), alphabetSize, alphabet, lengthPair.getValue(), meanCorrected));
}
}
/*
* In the empirical observations as described in paper:
* "The BOSS is concerned with time series classification in the presence of noise Patrick Schaefer" p.1519,
* showed that most of
* the time a alphabet size of 4 works best.
*/
public BOSSEnsembleClassifier(final Map<Integer, Integer> windowLengthsandWordLength, final double[] alphabet, final boolean meanCorrected) {
this(windowLengthsandWordLength, 4, alphabet, meanCorrected);
}
@Override
public Integer predict(final double[] univInstance) throws PredictionException {
HashMap<Integer, Integer> labelCount = new HashMap<>();
int votedLabel = 0;
int maxNumberOfVotes = Integer.MIN_VALUE;
for (BOSSClassifier boss : this.ensemble) {
Integer label = boss.predict(univInstance);
if (labelCount.containsKey(label)) {
labelCount.put(label, labelCount.get(label) + 1);
if (labelCount.get(label) > maxNumberOfVotes) {
votedLabel = label;
maxNumberOfVotes = labelCount.get(label);
}
} else {
labelCount.put(label, 1);
if (labelCount.get(label) > maxNumberOfVotes) {
votedLabel = label;
maxNumberOfVotes = labelCount.get(label);
}
}
}
return votedLabel;
}
@Override
public Integer predict(final List<double[]> multivInstance) throws PredictionException {
throw new UnsupportedOperationException("The BOSS-Esamble Classifier is an univirate classifier.");
}
@Override
public List<Integer> predict(final TimeSeriesDataset2 dataset) throws PredictionException {
ArrayList<Integer> predicts = new ArrayList<>();
for (double[][] matrix : dataset.getValueMatrices()) {
for (double[] instance : matrix) {
predicts.add(this.predict(instance));
}
}
return predicts;
}
@Override
public <U extends ASimplifiedTSClassifier<Integer>> ASimplifiedTSCLearningAlgorithm<Integer, U> getLearningAlgorithm(final TimeSeriesDataset2 dataset) {
throw new UnsupportedOperationException();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner/BOSSLearningAlgorithm.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import org.api4.java.algorithm.events.IAlgorithmEvent;
import ai.libs.jaicore.basic.IOwnerBasedAlgorithmConfig;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.filter.SFA;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.filter.SlidingWindowBuilder;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.filter.ZTransformer;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.util.HistogramBuilder;
/**
* @author Helen Beierling
* This class calculates all needed informations for the BOSS classifier. A fitted SFA and all
* histograms for the training samples.
*/
public class BOSSLearningAlgorithm extends ASimplifiedTSCLearningAlgorithm<Integer, BOSSClassifier> {
public interface IBossAlgorithmConfig extends IOwnerBasedAlgorithmConfig {
public static final String K_WINDOW_SIZE = "boss.windowsize";
public static final String K_ALPHABET_SIZE = "boss.alphabetsize";
public static final String K_ALPHABET = "boss.alphabet";
public static final String K_WORDLENGTH = "boss.wordlength";
public static final String K_MEANCORRECTED = "boss.meancorrected";
/**
* The size of the sliding window that is used over each instance and splits it into multiple
* smaller instances.
*/
@Key(K_WINDOW_SIZE)
public int windowSize();
/**
* The alphabet size determines the number of Bins for the SFA Histograms. Four was determined empirical
* as an optimal value for the alphabet size.
* cf.p. 1519 "The BOSS is concerned with time series classification in the presence of noise by Patrick Schäfer"
*
*/
@Key(K_ALPHABET_SIZE)
@DefaultValue("4")
public int alphabetSize();
/**
* The alphabet consists of doubles representing letters and defines each word.
*/
@Key(K_ALPHABET)
public double[] alphabet();
/**
* The word length determines the number of used DFT-coefficients. Where the DFT-coefficients are
* half the word length.
*/
@Key(K_WORDLENGTH)
public int wordLength();
/**
* If mean corrected is set to true than the first DFT coefficient is dropped to normalize the mean.
* c.f.p. 1519 "The BOSS is concerned with time series classification in the presence of noise by Patrick Schäfer"
*/
@Key(K_MEANCORRECTED)
public boolean meanCorrected();
}
/**
* The list contains the list of Histograms in which every matrix of the multivariate dataset results in.
*/
private List<ArrayList<HashMap<Integer, Integer>>> multivirateHistograms = new ArrayList<>();
/**
* Constians the histograms of one matrix for each instance one. Where the keys are the words which are double value
* sequences converted to an integer hash code and the values are the corresponding word counts.
*/
private ArrayList<HashMap<Integer, Integer>> histograms = new ArrayList<>();
// This class assumes that the optimal proportion of word length to window size is determined elsewhere and the corresponding
// drop of SFA words.
public BOSSLearningAlgorithm(final IBossAlgorithmConfig config, final BOSSClassifier classifier, final TimeSeriesDataset2 data) {
super(config, classifier, data);
}
@Override
public IAlgorithmEvent nextWithException() {
return null;
}
@Override
public BOSSClassifier call() {
this.multivirateHistograms.clear();
IBossAlgorithmConfig config = (IBossAlgorithmConfig) this.getConfig();
HistogramBuilder histoBuilder = new HistogramBuilder();
SFA sfa = new SFA(config.alphabet(), config.wordLength());
/* calculates the lookup table for the alphabet for the whole input dataset. */
SlidingWindowBuilder slide = new SlidingWindowBuilder();
slide.setDefaultWindowSize(config.windowSize());
TimeSeriesDataset2 data = this.getInput();
for (int matrix = 0; matrix < data.getNumberOfVariables(); matrix++) {
this.histograms.clear();
for (int instance = 0; instance < data.getNumberOfInstances(); instance++) {
/*
* Every instance results in an own histogram there for has its own HashMap of
* the the from key: word value: count of word.
*/
/*
* By the special fit transform an instance is transformed to a dataset. This
* is done because every instance creates a list of new smaller instances when
* split into sliding windows.
*/
TimeSeriesDataset2 tmp = slide.specialFitTransform(data.getValues(matrix)[instance]);
/* The from one instance resulting dataset is z-normalized. */
ZTransformer znorm = new ZTransformer();
for (int i = 0; i < tmp.getValues(0).length; i++) {
tmp.getValues(0)[i] = znorm.fitTransform(tmp.getValues(0)[i]);
}
// The SFA words for that dataset are computed using the precomputed MCB quantisation intervals
TimeSeriesDataset2 tmpTransformed = sfa.fitTransform(tmp);
// The occurring SFA words of the instance are getting counted with a parallel numerosity reduction.
Map<Integer, Integer> histogram = histoBuilder.histogramForInstance(tmpTransformed);
// Each instance in the dataset has its own histogram so the original dataset results in a list of histograms.
this.histograms.add(new HashMap<>(histogram));
}
// In the case of a multivariate dataset each matrix would have a list of histograms which than results
// in a list of lists of histograms.
// The Boss classifier however can not handle multivariate datasets.
this.multivirateHistograms.add(this.histograms);
}
// In the end all calculated and needed algortihms are set for the classifier.
BOSSClassifier model = this.getClassifier();
model.setTrainingData(this.getInput());
return model;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner/package-info.java
|
/**
* This package contains classifier implementations for time series classification problems.
*/
/**
* @author Julian
*
*/
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner;
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner/neighbors/NearestNeighborClassifier.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner.neighbors;
import java.util.ArrayList;
import java.util.Comparator;
import java.util.HashMap;
import java.util.List;
import java.util.Map.Entry;
import java.util.PriorityQueue;
import org.aeonbits.owner.ConfigCache;
import org.api4.java.ai.ml.core.exception.PredictionException;
import org.api4.java.common.metric.IDistanceMetric;
import ai.libs.jaicore.basic.IOwnerBasedAlgorithmConfig;
import ai.libs.jaicore.basic.sets.Pair;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner.ASimplifiedTSClassifier;
/**
* K-Nearest-Neighbor classifier for time series.
*
* Given an integer <code>k</code>, a distance measure <code>d</code> ({@link ai.libs.jaicore.ml.tsc.distances}), a training set of time series <code>TRAIN = {(x, y)}</code> and a test time series <code>T</code> (or a set of test time
* series).
* <p>
* The set of k nearest neighbors <code>NN</code> for <code>T</code> is a subset (or equal) of <code>TRAIN</code> with cardinality <code>k</code> such that for all <code>(T, S)</code> with <code>S</code> in <code>TRAIN\NN</code> holds
* <code>d(S, T) >= max_{T' in NN} d(S, T')</code>.
* </p>
* From the labels of the instances in <code>NN</code> the label for <code>T</code> is aggregated, e.g. via majority vote.
*
* @author fischor
*/
public class NearestNeighborClassifier extends ASimplifiedTSClassifier<Integer> {
/**
* Votes types that describe how to aggregate the prediciton for a test instance on its nearest neighbors found.
*/
public enum VoteType {
/**
* Majority vote with @see NearestNeighborClassifier#voteMajority.
*/
MAJORITY,
/**
* Weighted stepwise vote with @see NearestNeighborClassifier#voteWeightedStepwise.
*/
WEIGHTED_STEPWISE,
/**
* Weighted proportional to distance vote with @see NearestNeighborClassifier#voteWeightedProportionalToDistance.
*/
WEIGHTED_PROPORTIONAL_TO_DISTANCE,
}
/**
* Comparator class for the nearest neighbor priority queues, used for the nearest neighbor calculation. Sorts pairs of <code>(Integer: targetClass, Double: distance)</code> for nearest neigbors by distance ascending.
*/
private static class NearestNeighborComparator implements Comparator<Pair<Integer, Double>> {
@Override
public int compare(final Pair<Integer, Double> o1, final Pair<Integer, Double> o2) {
return -1 * o1.getY().compareTo(o2.getY());
}
}
/**
* Singleton comparator instance for the nearest neighbor priority queues, used for the nearest neighbor calculation.
*/
protected static final NearestNeighborComparator nearestNeighborComparator = new NearestNeighborComparator();
/** Number of neighbors. */
private int k;
/** Distance measure. */
private IDistanceMetric distanceMeasure;
/** Type of the voting. */
private VoteType voteType;
/** Value matrix containing the time series instances. Set by algorithm. */
protected double[][] values;
/**
* Timestamp matrix containing the timestamps of the instances. Set by the algorihm.
*/
protected double[][] timestamps;
/** Target values for the instances. Set by the algorithm. */
protected int[] targets;
/**
* Creates a k nearest neighbor classifier.
*
* @param k
* The number of nearest neighbors.
* @param distanceMeasure
* Distance measure for calculating the distances between every pair of train and test instances.
* @param voteType
* Vote type to use to aggregate the the classes of the the k nearest neighbors into a single class prediction.
*/
public NearestNeighborClassifier(final int k, final IDistanceMetric distanceMeasure, final VoteType voteType) {
// Parameter checks.
if (distanceMeasure == null) {
throw new IllegalArgumentException("Distance measure must not be null");
}
if (voteType == null) {
throw new IllegalArgumentException("Vote type must not be null.");
}
// Set attributes.
this.distanceMeasure = distanceMeasure;
this.k = k;
this.voteType = voteType;
}
/**
* Creates a k nearest neighbor classifier using majority vote.
*
* @param k
* The number of nearest neighbors.
* @param distanceMeasure
* Distance measure for calculating the distances between every pair of train and test instances.
*/
public NearestNeighborClassifier(final int k, final IDistanceMetric distanceMeasure) {
this(k, distanceMeasure, VoteType.MAJORITY);
}
/**
* Creates a 1 nearest neighbor classifier using majority vote.
*
* @param distanceMeasure
* Distance measure for calculating the distances between every pair of train and test instances.
*/
public NearestNeighborClassifier(final IDistanceMetric distanceMeasure) {
this(1, distanceMeasure, VoteType.MAJORITY);
}
/**
* Predicts on univariate instance.
*
* @param univInstance
* The univariate instance.
* @return Class prediction for the instance.
*/
@Override
public Integer predict(final double[] univInstance) throws PredictionException {
if (univInstance == null) {
throw new IllegalArgumentException("Instance to predict must not be null.");
}
return this.calculatePrediction(univInstance);
}
/**
* Predicts on a dataset.
*
* @param dataset
* The dataset.
* @return List of class predicitons for each instance of the dataset.
*/
@Override
public List<Integer> predict(final TimeSeriesDataset2 dataset) throws PredictionException {
double[][] testInstances = this.checkWhetherPredictionIsPossible(dataset);
// Calculate predictions.
ArrayList<Integer> predictions = new ArrayList<>(dataset.getNumberOfInstances());
for (double[] testInstance : testInstances) {
int prediction = this.calculatePrediction(testInstance);
predictions.add(prediction);
}
return predictions;
}
/**
* Calculates predicition on a single test instance.
*
* @param testInstance
* The test instance (not null assured within class).
* @return
*/
protected int calculatePrediction(final double[] testInstance) {
// Determine the k nearest neighbors for the test instance.
PriorityQueue<Pair<Integer, Double>> nearestNeighbors = this.calculateNearestNeigbors(testInstance);
// Vote on determined neighbors to create prediction and return prediction.
return this.vote(nearestNeighbors);
}
/**
* Determine the k nearest neighbors for a test instance.
*
* @param testInstance
* The time series to determine the k nearest neighbors for.
* @return Queue of the k nearest neighbors as pairs (class, distance).
*/
protected PriorityQueue<Pair<Integer, Double>> calculateNearestNeigbors(final double[] testInstance) {
int numberOfTrainInstances = this.values.length;
// Priority queue of (class, distance)-pairs for nearest neigbors, sorted by
// distance ascending.
PriorityQueue<Pair<Integer, Double>> nearestNeighbors = new PriorityQueue<>(nearestNeighborComparator);
// Calculate the k nearest neighbors.
for (int i = 0; i < numberOfTrainInstances; i++) {
double d = this.distanceMeasure.distance(testInstance, this.values[i]);
Pair<Integer, Double> neighbor = new Pair<>(this.targets[i], d);
nearestNeighbors.add(neighbor);
if (nearestNeighbors.size() > this.k) {
nearestNeighbors.poll();
}
}
return nearestNeighbors;
}
/**
* Performs a vote on the nearest neighbors found. Delegates the vote according to the vote type.
*
* @param nearestNeighbors
* Priority queue of (class, distance)-pairs for nearest neigbors, sorted by distance ascending. (Not null assured within class)
* @return Result of the vote, i.e. the predicted class.
*/
protected int vote(final PriorityQueue<Pair<Integer, Double>> nearestNeighbors) {
switch (this.voteType) {
case WEIGHTED_STEPWISE:
return this.voteWeightedStepwise(nearestNeighbors);
case WEIGHTED_PROPORTIONAL_TO_DISTANCE:
return this.voteWeightedProportionalToDistance(nearestNeighbors);
case MAJORITY:
default:
return this.voteMajority(nearestNeighbors);
}
}
/**
* Performs a vote with stepwise weights 1, 2, .., k on the set nearest neighbors found.
*
* @param nearestNeighbors
* Priority queue of (class, distance)-pairs for nearest neigbors, sorted by distance ascending. (Not null assured within class)
* @return Result of the vote, i.e. the predicted class.
*/
protected int voteWeightedStepwise(final PriorityQueue<Pair<Integer, Double>> nearestNeighbors) {
// Voting.
HashMap<Integer, Integer> votes = new HashMap<>();
int weight = 1;
while (!nearestNeighbors.isEmpty()) {
Pair<Integer, Double> neighbor = nearestNeighbors.poll();
Integer targetClass = neighbor.getX();
Integer currentVotesOnTargetClass = votes.get(targetClass);
if (currentVotesOnTargetClass == null) {
votes.put(targetClass, weight);
} else {
votes.put(targetClass, currentVotesOnTargetClass + weight);
}
weight++;
}
// Return most voted target (class that got most weights).
Integer maxWeightOfVotes = Integer.MIN_VALUE;
Integer mostVotedTargetClass = -1;
for (Entry<Integer, Integer> entry : votes.entrySet()) {
int targetClass = entry.getKey();
int votedWeightsForTargetClass = entry.getValue();
if (votedWeightsForTargetClass > maxWeightOfVotes) {
maxWeightOfVotes = votedWeightsForTargetClass;
mostVotedTargetClass = targetClass;
}
}
return mostVotedTargetClass;
}
/**
* Performs a vote with weights proportional to the distance on the set nearest neighbors found.
*
* @param nearestNeighbors
* Priority queue of (class, distance)-pairs for nearest neigbors, sorted by distance ascending. (Not null assured within class)
* @return Result of the vote, i.e. the predicted class.
*/
protected int voteWeightedProportionalToDistance(final PriorityQueue<Pair<Integer, Double>> nearestNeighbors) {
// Voting.
HashMap<Integer, Double> votes = new HashMap<>();
for (Pair<Integer, Double> neighbor : nearestNeighbors) {
Integer targetClass = neighbor.getX();
double distance = neighbor.getY();
Double currentVotesOnTargetClass = votes.get(targetClass);
if (currentVotesOnTargetClass == null) {
votes.put(targetClass, 1.0 / distance);
} else {
votes.put(targetClass, currentVotesOnTargetClass + 1.0 / distance);
}
}
// Return most voted target (class that got most weights).
Double maxWeightOfVotes = Double.MIN_VALUE;
Integer mostVotedTargetClass = -1;
for (Entry<Integer, Double> entry : votes.entrySet()) {
int targetClass = entry.getKey();
double votedWeightsForTargetClass = entry.getValue();
if (votedWeightsForTargetClass > maxWeightOfVotes) {
maxWeightOfVotes = votedWeightsForTargetClass;
mostVotedTargetClass = targetClass;
}
}
return mostVotedTargetClass;
}
/**
* Performs a majority vote on the set nearest neighbors found.
*
* @param nearestNeighbors
* Priority queue of (class, distance)-pairs for nearest neigbors, sorted by distance ascending. (Not null assured within class)
* @return Result of the vote, i.e. the predicted class.
*/
protected int voteMajority(final PriorityQueue<Pair<Integer, Double>> nearestNeighbors) {
// Voting.
HashMap<Integer, Integer> votes = new HashMap<>();
for (Pair<Integer, Double> neighbor : nearestNeighbors) {
Integer targetClass = neighbor.getX();
Integer currentVotesOnTargetClass = votes.get(targetClass);
if (currentVotesOnTargetClass == null) {
votes.put(targetClass, 1);
} else {
votes.put(targetClass, currentVotesOnTargetClass + 1);
}
}
// Return most voted target.
Integer maxNumberOfVotes = Integer.MIN_VALUE;
Integer mostVotedTargetClass = -1;
for (Entry<Integer, Integer> entry : votes.entrySet()) {
int targetClass = entry.getKey();
int numberOfVotesForTargetClass = entry.getValue();
if (numberOfVotesForTargetClass > maxNumberOfVotes) {
maxNumberOfVotes = numberOfVotesForTargetClass;
mostVotedTargetClass = targetClass;
}
}
return mostVotedTargetClass;
}
/**
* Sets the value matrix.
*
* @param values
*/
protected void setValues(final double[][] values) {
if (values == null) {
throw new IllegalArgumentException("Values must not be null");
}
this.values = values;
}
/**
* Sets the timestamps.
*
* @param timestamps
*/
protected void setTimestamps(final double[][] timestamps) {
this.timestamps = timestamps;
}
/**
* Sets the targets.
*
* @param targets
*/
protected void setTargets(final int[] targets) {
if (targets == null) {
throw new IllegalArgumentException("Targets must not be null");
}
this.targets = targets;
}
/**
* Getter for the k value, @see #k.
*
* @return k
*/
public int getK() {
return this.k;
}
/**
* Getter for the vote type. @see #voteType.
*
* @return The vote type.
*/
public VoteType getVoteType() {
return this.voteType;
}
/**
* Getter for the distance measure. @see #distanceMeasure.
*
* @return
*/
public IDistanceMetric getDistanceMeasure() {
return this.distanceMeasure;
}
@Override
public NearestNeighborLearningAlgorithm getLearningAlgorithm(final TimeSeriesDataset2 dataset) {
return new NearestNeighborLearningAlgorithm(ConfigCache.getOrCreate(IOwnerBasedAlgorithmConfig.class), this, dataset);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner/neighbors/NearestNeighborLearningAlgorithm.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner.neighbors;
import org.api4.java.algorithm.events.IAlgorithmEvent;
import org.api4.java.algorithm.exceptions.AlgorithmException;
import ai.libs.jaicore.basic.IOwnerBasedAlgorithmConfig;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner.ASimplifiedTSCLearningAlgorithm;
/**
* Training algorithm for the nearest neighbors classifier.
*
* This algorithm just delegates the value matrix, timestamps and targets to the
* classifier.
*
* @author fischor
*/
public class NearestNeighborLearningAlgorithm extends ASimplifiedTSCLearningAlgorithm<Integer, NearestNeighborClassifier> {
protected NearestNeighborLearningAlgorithm(final IOwnerBasedAlgorithmConfig config, final NearestNeighborClassifier classifier, final TimeSeriesDataset2 input) {
super(config, classifier, input);
}
@Override
public NearestNeighborClassifier call() throws AlgorithmException {
TimeSeriesDataset2 dataset = this.getInput();
if (dataset == null) {
throw new AlgorithmException("No input data set.");
}
if (dataset.isMultivariate()) {
throw new UnsupportedOperationException("Multivariate datasets are not supported.");
}
// Retrieve data from dataset.
double[][] values = dataset.getValuesOrNull(0);
// Check data.
if (values == null) {
throw new AlgorithmException("Empty input data set.");
}
int[] targets = dataset.getTargets();
if (targets == null) {
throw new AlgorithmException("Empty targets.");
}
// Update model.
NearestNeighborClassifier model = this.getClassifier();
model.setValues(values);
model.setTimestamps(dataset.getTimestampsOrNull(0));
model.setTargets(targets);
return model;
}
@Override
public IAlgorithmEvent nextWithException() {
throw new UnsupportedOperationException();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner/neighbors/ShotgunEnsembleClassifier.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner.neighbors;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Map.Entry;
import org.aeonbits.owner.ConfigCache;
import org.api4.java.ai.ml.core.exception.PredictionException;
import org.api4.java.common.metric.IDistanceMetric;
import ai.libs.jaicore.basic.metric.ShotgunDistance;
import ai.libs.jaicore.basic.sets.Pair;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner.ASimplifiedTSClassifier;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner.neighbors.ShotgunEnsembleLearnerAlgorithm.IShotgunEnsembleLearnerConfig;
/**
* Implementation of Shotgun Ensemble Classifier as published in "Towards Time
* Series Classfication without Human Preprocessing" by Patrick Schäfer (2014).
*
* The Shotgun Classifier is based 1-NN and the Shotgun Distance.
*
* The Shotgun Ensemble Algorithm {@link ShotgunEnsembleAlgoritm} determines for
* specific window lengths the number of correct predicitions on the training
* data using the leave-one-out technique. The <code>bestScore</code> is the
* highest number of correct predicitions over all window lengths. Given a
* <code>factor</code> in <code>(0,1]</code>, the window lengths where
* <code>correct * factor > bestScore</code> are used in an ensemble of Shotgun
* Classifiers to create an overall predicition.
*
* @author fischor
*/
public class ShotgunEnsembleClassifier extends ASimplifiedTSClassifier<Integer> {
/**
* Factor used to determine whether or not to include a window length into the
* overall predicition.
*/
protected double factor;
/** Value matrix containing the time series instances. Set by algorithm. */
protected double[][] values;
/** Target values for the instances. Set by the algorithm. */
protected int[] targets;
/**
* The nearest neighbor classifier used for prediction. Set by the algorithm.
*/
protected NearestNeighborClassifier nearestNeighborClassifier;
/**
* The Shotgun Distance used by the {@link #nearestNeighborClassifier}. Set by
* the algorithm.
*/
protected ShotgunDistance shotgunDistance;
/**
* Holds pairs of (number of correct predictions, window length) obtained in
* training phase.
*/
protected ArrayList<Pair<Integer, Integer>> windows;
/**
* The best score. States the highest number of correct predicitions for every
* window length used in training phase (leave-one-out).
*/
protected int bestScore;
private final IShotgunEnsembleLearnerConfig config;
/**
* Creates a Shotgun Ensemble classifier.
*
* @param algorithm The training algorithm.
* @param factor Factor used to determine whether or not to include a window
* length into the overall predicition.
*/
public ShotgunEnsembleClassifier(final int minWindowLength, final int maxWindowLength, final boolean meanNormalization, final double factor) {
super();
this.config = ConfigCache.getOrCreate(IShotgunEnsembleLearnerConfig.class);
if (minWindowLength < 1) {
throw new IllegalArgumentException("The parameter minWindowLength must be greater equal to 1.");
}
if (maxWindowLength < 1) {
throw new IllegalArgumentException("The parameter maxWindowLength must be greater equal to 1.");
}
if (minWindowLength > maxWindowLength) {
throw new IllegalAccessError("The parameter maxWindowsLength must be greater equal to parameter minWindowLength");
}
this.config.setProperty(IShotgunEnsembleLearnerConfig.K_WINDOWLENGTH_MIN, "" + minWindowLength);
this.config.setProperty(IShotgunEnsembleLearnerConfig.K_WINDOWLENGTH_MAX, "" + maxWindowLength);
this.config.setProperty(IShotgunEnsembleLearnerConfig.K_MEANNORMALIZATION, "" + meanNormalization);
if ((factor <= 0) || (factor > 1)) {
throw new IllegalArgumentException("The parameter factor must be in (0,1]");
}
this.factor = factor;
}
/**
* Calculates predicitions for a test instance using 1NN with Shotgun Distance
* and different window lengths.
*
* @param testInstance The test instance.
* @return Map of (window length, prediciton) pairs.
* @throws PredictionException
*/
protected Map<Integer, Integer> calculateWindowLengthPredictions(final double[] testInstance) throws PredictionException {
// Map holding (windowLength, predicition for instance) pairs.
Map<Integer, Integer> windowLengthPredicitions = new HashMap<>();
for (Pair<Integer, Integer> window : this.windows) {
int correct = window.getX();
int windowLength = window.getY();
this.shotgunDistance.setWindowLength(windowLength);
if (correct > this.bestScore * this.factor) {
int prediction = this.nearestNeighborClassifier.predict(testInstance);
windowLengthPredicitions.put(windowLength, prediction);
}
}
return windowLengthPredicitions;
}
/**
* Returns the most frequent predicition given a Map of (window length,
* prediciton) pairs.
*
* @param windowLengthPredicitions Map of (window length, prediciton) pairs.
* @return The most frequent predicition.
*/
protected Integer mostFrequentLabelFromWindowLengthPredicitions(final Map<Integer, Integer> windowLengthPredicitions) {
// Count frequency for labels.
Map<Integer, Integer> labelFrequencyMap = new HashMap<>();
for (Integer label : windowLengthPredicitions.values()) {
if (labelFrequencyMap.containsKey(label)) {
labelFrequencyMap.put(label, labelFrequencyMap.get(label) + 1);
} else {
labelFrequencyMap.put(label, 1);
}
}
// Return most frequent label.
int topFrequency = -1;
int mostFrequentLabel = 0;
for (Entry<Integer, Integer> entry : labelFrequencyMap.entrySet()) {
int label = entry.getKey();
int labelFrequency = entry.getValue();
if (labelFrequency > topFrequency) {
topFrequency = labelFrequency;
mostFrequentLabel = label;
}
}
return mostFrequentLabel;
}
/**
* Calculates predicitions for a test dataset using 1NN with Shotgun Distance
* and different window lengths.
*
* @param dataset The dataset to predict for.
* @return Map of (window length, predicitions) pairs.
* @throws PredictionException
*/
protected Map<Integer, List<Integer>> calculateWindowLengthPredictions(final TimeSeriesDataset2 dataset) throws PredictionException {
// Map holding (windowLength, prediction for dataset) pairs.
Map<Integer, List<Integer>> windowLengthPredicitions = new HashMap<>();
for (Pair<Integer, Integer> window : this.windows) {
int correct = window.getX();
int windowLength = window.getY();
this.shotgunDistance.setWindowLength(windowLength);
if (correct > this.bestScore * this.factor) {
List<Integer> predictions = this.nearestNeighborClassifier.predict(dataset);
windowLengthPredicitions.put(windowLength, predictions);
}
}
return windowLengthPredicitions;
}
/**
* Returns for each instance the most frequent predicitions as contained in a
* Map of (window length, list of prediciton for each instance) pairs.
*
* @param windowLengthPredicitions Map of (window length, list of prediciton for
* each instance) pairs.
* @return The most frequent predicition for each instace.
*/
protected List<Integer> mostFrequentLabelsFromWindowLengthPredicitions(final Map<Integer, List<Integer>> windowLengthPredicitions) {
// Return most frequent label for each instance.
int numberOfInstances = windowLengthPredicitions.values().iterator().next().size();
List<Integer> predicitions = new ArrayList<>(numberOfInstances);
for (int i = 0; i < numberOfInstances; i++) {
// Map holding (windowLength, predicition for instance) pairs.
Map<Integer, Integer> windowLabelsForInstance = new HashMap<>();
for (Entry<Integer, List<Integer>> entry : windowLengthPredicitions.entrySet()) {
int windowLength = entry.getKey();
int predictionForWindowLength = entry.getValue().get(i);
windowLabelsForInstance.put(windowLength, predictionForWindowLength);
}
int mostFrequentLabelForInstance = this.mostFrequentLabelFromWindowLengthPredicitions(windowLabelsForInstance);
predicitions.add(mostFrequentLabelForInstance);
}
return predicitions;
}
/**
* Predicts on univariate instance.
*
* @param univInstance The univariate instance.
* @return Class prediction for the instance.
*/
@Override
public Integer predict(final double[] univInstance) throws PredictionException {
if (univInstance == null) {
throw new IllegalArgumentException("Instance to predict must not be null.");
}
Map<Integer, Integer> windowLengthPredicitions = this.calculateWindowLengthPredictions(univInstance);
return this.mostFrequentLabelFromWindowLengthPredicitions(windowLengthPredicitions);
}
/**
* Predicts on a dataset.
*
* @param dataset The dataset.
* @return List of class predicitons for each instance of the dataset.
*/
@Override
public List<Integer> predict(final TimeSeriesDataset2 dataset) throws PredictionException {
this.checkWhetherPredictionIsPossible(dataset);
Map<Integer, List<Integer>> windowLengthPredicitions = this.calculateWindowLengthPredictions(dataset);
return this.mostFrequentLabelsFromWindowLengthPredicitions(windowLengthPredicitions);
}
/**
* Sets the value matrix.
*
* @param values
*/
protected void setValues(final double[][] values) {
if (values == null) {
throw new IllegalArgumentException("Values must not be null");
}
this.values = values;
}
/**
* Sets the targets.
*
* @param targets
*/
protected void setTargets(final int[] targets) {
if (targets == null) {
throw new IllegalArgumentException("Targets must not be null");
}
this.targets = targets;
}
/**
* Sets the windows and also retreives and sets the @see #bestScore from these
* windows.
*
* @param windows @see #windows
*/
protected void setWindows(final ArrayList<Pair<Integer, Integer>> windows) {
this.windows = windows;
// Best score.
int tBestScore = -1;
for (Pair<Integer, Integer> window : windows) {
int correct = window.getX();
if (correct > tBestScore) {
tBestScore = correct;
}
}
this.bestScore = tBestScore;
}
/**
* Sets the nearest neighbor classifier, {@link #nearestNeighborClassifier}.
*
* @param nearestNeighborClassifier
*/
protected void setNearestNeighborClassifier(final NearestNeighborClassifier nearestNeighborClassifier) {
IDistanceMetric distanceMeasure = nearestNeighborClassifier.getDistanceMeasure();
if (!(distanceMeasure instanceof ShotgunDistance)) {
throw new IllegalArgumentException("The nearest neighbor classifier must use a ShotgunDistance as dsitance measure.");
} else {
this.shotgunDistance = (ShotgunDistance) distanceMeasure;
}
this.nearestNeighborClassifier = nearestNeighborClassifier;
}
@Override
public ShotgunEnsembleLearnerAlgorithm getLearningAlgorithm(final TimeSeriesDataset2 dataset) {
return new ShotgunEnsembleLearnerAlgorithm(this.config, this, dataset);
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/learner/neighbors/ShotgunEnsembleLearnerAlgorithm.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner.neighbors;
import java.util.ArrayList;
import org.api4.java.algorithm.events.IAlgorithmEvent;
import org.api4.java.algorithm.exceptions.AlgorithmException;
import ai.libs.jaicore.basic.IOwnerBasedAlgorithmConfig;
import ai.libs.jaicore.basic.metric.ShotgunDistance;
import ai.libs.jaicore.basic.sets.Pair;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.learner.ASimplifiedTSCLearningAlgorithm;
/**
* Implementation of Shotgun Ensemble Algorihm as published in "Towards Time
* Series Classfication without Human Preprocessing" by Patrick Schäfer (2014).
*
* Given a maximal window length <code>maxWindowLength</code> and a minumum
* window length <code>minWindowLength</code>, the Shotgun Ensemble algorithm
* determines for each of the window lengths form <code>maxWindowLength</code>
* downto <code>minWindowLength</code> the number of correct predicitions on the
* training data using the leave-one-out technique.
*
* @author fischor
*/
public class ShotgunEnsembleLearnerAlgorithm extends ASimplifiedTSCLearningAlgorithm<Integer, ShotgunEnsembleClassifier> {
public interface IShotgunEnsembleLearnerConfig extends IOwnerBasedAlgorithmConfig {
public static final String K_WINDOWLENGTH_MIN = "windowlength.min";
public static final String K_WINDOWLENGTH_MAX = "windowlength.max";
public static final String K_MEANNORMALIZATION = "meannormalization";
@Key(K_WINDOWLENGTH_MIN)
public int windowSizeMin();
@Key(K_WINDOWLENGTH_MAX)
public int windowSizeMax();
@Key(K_MEANNORMALIZATION)
@DefaultValue("false")
public boolean meanNormalization();
}
public ShotgunEnsembleLearnerAlgorithm(final IShotgunEnsembleLearnerConfig config, final ShotgunEnsembleClassifier classifier, final TimeSeriesDataset2 dataset) {
super(config, classifier, dataset);
}
@Override
public IAlgorithmEvent nextWithException() {
throw new UnsupportedOperationException();
}
@Override
public IShotgunEnsembleLearnerConfig getConfig() {
return (IShotgunEnsembleLearnerConfig) super.getConfig();
}
@Override
public ShotgunEnsembleClassifier call() throws AlgorithmException {
TimeSeriesDataset2 dataset = this.getInput();
if (dataset == null) {
throw new AlgorithmException("No input data set.");
}
if (dataset.isMultivariate()) {
throw new UnsupportedOperationException("Multivariate datasets are not supported.");
}
// Retrieve data from dataset.
double[][] values = dataset.getValuesOrNull(0);
// Check data.
if (values == null) {
throw new AlgorithmException("Empty input data set.");
}
int[] targets = dataset.getTargets();
if (targets == null) {
throw new AlgorithmException("Empty targets.");
}
// Holds pairs of (number of correct predictions, window length).
ArrayList<Pair<Integer, Integer>> scores = new ArrayList<>();
for (int windowLength = this.getConfig().windowSizeMax(); windowLength >= this.getConfig().windowSizeMin(); windowLength--) {
int correct = 0;
// 1-NN with Leave-One-Out CV.
ShotgunDistance shotgunDistance = new ShotgunDistance(windowLength, this.getConfig().meanNormalization());
for (int i = 0; i < values.length; i++) {
// Predict for i-th instance.
double minDistance = Double.MAX_VALUE;
int instanceThatMinimizesDistance = -1;
for (int j = 0; j < values.length; j++) {
if (i != j) {
double distance = shotgunDistance.distance(values[i], values[j]);
if (distance < minDistance) {
minDistance = distance;
instanceThatMinimizesDistance = j;
}
}
}
// Check, if Leave-One-Out prediction for i-th was correct.
if (targets[i] == targets[instanceThatMinimizesDistance]) {
correct++;
}
}
scores.add(new Pair<>(correct, windowLength));
}
// Update model.
NearestNeighborClassifier nearestNeighborClassifier = new NearestNeighborClassifier(new ShotgunDistance(this.getConfig().windowSizeMax(), this.getConfig().meanNormalization()));
try {
nearestNeighborClassifier.train(dataset);
} catch (Exception e) {
throw new AlgorithmException("Cant train nearest neighbor classifier.", e);
}
ShotgunEnsembleClassifier model = this.getClassifier();
model.setWindows(scores);
model.setNearestNeighborClassifier(nearestNeighborClassifier);
return model;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/model/INDArrayTimeseries.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.model;
import org.api4.java.ai.ml.core.dataset.schema.attribute.ITimeseries;
import org.nd4j.linalg.api.ndarray.INDArray;
public interface INDArrayTimeseries extends ITimeseries<INDArray> {
public int length();
public double[] getPoint();
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/model/NDArrayTimeseries.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.model;
import org.nd4j.linalg.api.ndarray.INDArray;
import ai.libs.jaicore.basic.sets.ElementDecorator;
public class NDArrayTimeseries extends ElementDecorator<INDArray> implements INDArrayTimeseries {
public NDArrayTimeseries(final INDArray element) {
super(element);
}
@Override
public INDArray getValue() {
return this.getElement();
}
@Override
public int length() {
return (int) this.getElement().length();
}
@Override
public double[] getPoint() {
return this.getElement().toDoubleVector();
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/model/package-info.java
|
/**
*
*/
/**
* @author mwever
*
*/
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.model;
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/quality/FStat.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.quality;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map.Entry;
/**
* F-Stat quality measure performing a analysis of variance according to chapter
* 3.2 of the original paper. It analyzes the ratio of the variability between
* the group of instances within a class to the variability within the class
* groups.
*
* @author Julian Lienen
*
*/
public class FStat implements IQualityMeasure {
/**
* Generated serial version UID.
*/
private static final long serialVersionUID = 6991529180002046551L;
/**
* {@inheritDoc}
*/
@Override
public double assessQuality(final List<Double> distances, final int[] classValues) {
// Order class distances
HashMap<Integer, List<Double>> classDistances = new HashMap<>();
for (int i = 0; i < distances.size(); i++) {
if (!classDistances.containsKey(classValues[i])) {
classDistances.put(classValues[i], new ArrayList<>());
}
classDistances.get(classValues[i]).add(distances.get(i));
}
int numClasses = classDistances.size();
// Calculate class and overall means
HashMap<Integer, Double> classMeans = new HashMap<>();
for (Entry<Integer, List<Double>> entry : classDistances.entrySet()) {
Integer clazz = entry.getKey();
classMeans.put(clazz, entry.getValue().stream().mapToDouble(a -> a).average().getAsDouble());
}
double completeMean = distances.stream().mapToDouble(a -> a).average().getAsDouble();
double denominator = 0;
// Calculate actual F score
double result = 0;
for (Entry<Integer, Double> entry : classMeans.entrySet()) {
Integer clazz = entry.getKey();
double mean = entry.getValue();
result += Math.pow(mean - completeMean, 2);
for (Double dist : classDistances.get(clazz)) {
denominator += Math.pow(dist - mean, 2);
}
}
result /= numClasses - 1;
denominator /= distances.size() - numClasses;
if (denominator == 0) {
throw new IllegalArgumentException("Given arguments yield a 0 " + denominator);
}
result /= denominator;
return result;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/quality/IQualityMeasure.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.quality;
import java.io.Serializable;
import java.util.List;
/**
* Interface for a quality measure assessing distances of instances to a
* shapelet given the corresponding class values. This functional interface is
* used within the Shapelet Transform approach to assess shapelet candidates.
*
* @author Julian Lienen
*
*/
public interface IQualityMeasure extends Serializable {
/**
* Computes a quality score based on the distances of each instance to the
* shapelet and the corresponding <code>classValues</code>.
*
* @param distances
* List of distances storing the distance of each instance to a
* shapelet
* @param classValues
* The class values of the instances
* @return Returns the calculated quality score
*/
public double assessQuality(final List<Double> distances, final int[] classValues);
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/shapelets/Shapelet.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.shapelets;
import java.util.Arrays;
import java.util.Collections;
import java.util.List;
/**
* Implementation of a shapelet, i. e. a specific subsequence of a time series
* representing a characteristic shape.
*
* @author Julian Lienen
*
*/
public class Shapelet {
/**
* The data vector of the shapelet.
*/
private double[] data;
/**
* The start index of the shapelet in the origin time series.
*/
private int startIndex;
/**
* The length of the shapelet.
*/
private int length;
/**
* The instance index which is assigned when extracting the shapelet from a
* given time series.
*/
private int instanceIndex;
/**
* The quality determined by an assessment function.
*/
private double determinedQuality;
/**
* Constructs a shapelet specified by the given parameters.
*
* @param data
* See {@link Shapelet#data}
* @param startIndex
* See {@link Shapelet#startIndex}
* @param length
* See {@link Shapelet#length}
* @param instanceIndex
* See {@link Shapelet#instanceIndex}
* @param determinedQuality
* See {@link Shapelet#determinedQuality}
*/
public Shapelet(final double[] data, final int startIndex, final int length, final int instanceIndex, final double determinedQuality) {
this.data = data;
this.startIndex = startIndex;
this.length = length;
this.instanceIndex = instanceIndex;
this.determinedQuality = determinedQuality;
}
/**
* Constructs a shapelet specified by the given parameters.
*
* @param data
* See {@link Shapelet#data}
* @param startIndex
* See {@link Shapelet#startIndex}
* @param length
* See {@link Shapelet#length}
* @param instanceIndex
* See {@link Shapelet#instanceIndex}
*/
public Shapelet(final double[] data, final int startIndex, final int length, final int instanceIndex) {
this.data = data;
this.startIndex = startIndex;
this.length = length;
this.instanceIndex = instanceIndex;
}
/**
* Getter for {@link Shapelet#data}.
*
* @return Return the shapelet's data vector
*/
public double[] getData() {
return this.data;
}
/**
* Getter for {@link Shapelet#length}.
*
* @return Returns the shapelet's length
*/
public int getLength() {
return this.length;
}
/**
* Getter for {@link Shapelet#startIndex}.
*
* @return Returns the shapelet's start index.
*/
public int getStartIndex() {
return this.startIndex;
}
/**
* Getter for {@link Shapelet#instanceIndex}.
*
* @return Returns the shapelet's instance index.
*/
public int getInstanceIndex() {
return this.instanceIndex;
}
/**
* Getter for {@link Shapelet#determinedQuality}.
*
* @return Returns the shapelet's determined quality.
*/
public double getDeterminedQuality() {
return this.determinedQuality;
}
/**
* Setter for {@link Shapelet#determinedQuality}.
*
* @param determinedQuality
* The new value to be set
*/
public void setDeterminedQuality(final double determinedQuality) {
this.determinedQuality = determinedQuality;
}
@Override
public int hashCode() {
final int prime = 31;
int result = 1;
result = prime * result + Arrays.hashCode(this.data);
long temp;
temp = Double.doubleToLongBits(this.determinedQuality);
result = prime * result + (int) (temp ^ (temp >>> 32));
result = prime * result + this.instanceIndex;
result = prime * result + this.length;
result = prime * result + this.startIndex;
return result;
}
/**
* {@inheritDoc}
*/
@Override
public boolean equals(final Object obj) {
if (obj instanceof Shapelet) {
Shapelet other = (Shapelet) obj;
if (this.data == null && other.getData() != null || this.data != null && other.getData() == null) {
return false;
}
return (this.data == null && other.getData() == null || Arrays.equals(this.data, other.getData())) && this.length == other.getLength() && this.determinedQuality == other.determinedQuality
&& this.instanceIndex == other.instanceIndex;
}
return super.equals(obj);
}
/**
* {@inheritDoc}
*/
@Override
public String toString() {
return "Shapelet [data=" + Arrays.toString(this.data) + ", startIndex=" + this.startIndex + ", length=" + this.length + ", instanceIndex=" + this.instanceIndex + ", determinedQuality=" + this.determinedQuality + "]";
}
/**
* Function sorting a list of shapelets in place by the length (ascending).
*
* @param shapelets
* The list to be sorted in place.
*/
public static void sortByLengthAsc(final List<Shapelet> shapelets) {
shapelets.sort((s1, s2) -> Integer.compare(s1.getLength(), s2.getLength()));
}
/**
* Returns the shapelet with the highest quality in the given list
* <code>shapelets</code>.
*
* @param shapelets
* The list of shapelets which is evaluated
* @return Returns the shapelet with the highest determined quality
*/
public static Shapelet getHighestQualityShapeletInList(final List<Shapelet> shapelets) {
return Collections.max(shapelets, (s1, s2) -> (-1) * Double.compare(s1.getDeterminedQuality(), s2.getDeterminedQuality()));
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/shapelets
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/shapelets/search/AMinimumDistanceSearchStrategy.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.shapelets.search;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.shapelets.Shapelet;
/**
* Abstract class for minimum distance search strategies. Subclasses implement
* functionality to find the minimum distance between a given {@link Shapelet}
* object and a time series.
*
* @author Julian Lienen
*
*/
public abstract class AMinimumDistanceSearchStrategy {
/**
* Indicator whether Bessel's correction should be used within any distance
* calculation;
*/
protected boolean useBiasCorrection;
/**
* Constructor.
*
* @param useBiasCorrection
* See {@link AMinimumDistanceSearchStrategy#useBiasCorrection}
*/
public AMinimumDistanceSearchStrategy(final boolean useBiasCorrection) {
this.useBiasCorrection = useBiasCorrection;
}
/**
* Function returning the minimum distance among all subsequences of the given
* <code>timeSeries</code> to the <code>shapelet</code>'s data.
*
* @param shapelet
* The shapelet to be compared to all subsequences
* @param timeSeries
* The time series which subsequences are compared to the shapelet's
* data
* @return Return the minimum distance among all subsequences
*/
public abstract double findMinimumDistance(final Shapelet shapelet, final double[] timeSeries);
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/shapelets
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/shapelets/search/EarlyAbandonMinimumDistanceSearchStrategy.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.shapelets.search;
import java.util.List;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.shapelets.Shapelet;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.util.MathUtil;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.util.TimeSeriesUtil;
/**
* Class implementing a search strategy used for finding the minimum distance of
* a {@link Shapelet} object to a time series. The approach uses early
* abandoning as described in algorithm 2 in the paper 'Jason Lines, Luke M.
* Davis, Jon Hills, and Anthony Bagnall. 2012. A shapelet transform for time
* series classification. In Proceedings of the 18th ACM SIGKDD international
* conference on Knowledge discovery and data mining (KDD '12). ACM, New York,
* NY, USA, 289-297.'.
*
* @author Julian Lienen
*
*/
public class EarlyAbandonMinimumDistanceSearchStrategy extends AMinimumDistanceSearchStrategy {
/**
* Standard constructor.
*
* @param useBiasCorrection
* See {@link AMinimumDistanceSearchStrategy#useBiasCorrection}
*/
public EarlyAbandonMinimumDistanceSearchStrategy(final boolean useBiasCorrection) {
super(useBiasCorrection);
}
/**
* Optimized function returning the minimum distance among all subsequences of
* the given <code>timeSeries</code> to the <code>shapelet</code>'s data. This
* function implements the algorithm 2 mentioned in the original paper. It
* performs the similarity search with online normalization and early abandon.
*
* @param shapelet
* The shapelet to be compared to all subsequences
* @param timeSeries
* The time series which subsequences are compared to the shapelet's
* data
* @return Return the minimum distance among all subsequences
*/
@Override
public double findMinimumDistance(final Shapelet shapelet, final double[] timeSeries) {
double length = shapelet.getLength();
int m = timeSeries.length;
// Order normalized shapelet values
final double[] sPrimeVector = shapelet.getData();
final List<Integer> aVector = TimeSeriesUtil.sortIndexes(sPrimeVector, false); // descending
final double[] fVector = TimeSeriesUtil.zNormalize(TimeSeriesUtil.getInterval(timeSeries, 0, shapelet.getLength()), this.useBiasCorrection);
// Online normalization
double p = 0;
double q = 0;
p = MathUtil.sum(TimeSeriesUtil.getInterval(timeSeries, 0, shapelet.getLength()));
for (int i = 0; i < length; i++) {
q += timeSeries[i] * timeSeries[i];
}
double b = MathUtil.singleSquaredEuclideanDistance(sPrimeVector, fVector);
for (int i = 1; i <= m - length; i++) {
double ti = timeSeries[i - 1];
double til = timeSeries[i - 1 + shapelet.getLength()];
p -= ti;
q -= ti * ti;
p += til;
q += til * til;
double xBar = p / length;
double s = q / (length) - xBar * xBar;
if (s < 0.000000001d) {
s = 0d;
} else {
s = Math.sqrt((this.useBiasCorrection ? (length / (length - 1d)) : 1d) * s);
}
int j = 0;
double d = 0d;
// Early abandon
while (j < length && d < b) {
final double normVal = (s == 0.0 ? 0d : (timeSeries[i + aVector.get(j)] - xBar) / s);
final double diff = sPrimeVector[aVector.get(j)] - normVal;
d += diff * diff;
j++;
}
if (j == length && d < b) {
b = d;
}
}
return b / length;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/shapelets
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/shapelets/search/ExhaustiveMinimumDistanceSearchStrategy.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.shapelets.search;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.shapelets.Shapelet;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.util.MathUtil;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.util.TimeSeriesUtil;
/**
* Class implementing a search strategy used for finding the minimum distance of
* a {@link Shapelet} object to a time series. The approach uses an exhaustive
* search as described in the paper 'Jason Lines, Luke M. Davis, Jon Hills, and
* Anthony Bagnall. 2012. A shapelet transform for time series classification.
* In Proceedings of the 18th ACM SIGKDD international conference on Knowledge
* discovery and data mining (KDD '12). ACM, New York, NY, USA, 289-297.'.
*
* @author Julian Lienen
*
*/
public class ExhaustiveMinimumDistanceSearchStrategy extends AMinimumDistanceSearchStrategy {
/**
* Standard constructor.
*
* @param useBiasCorrection
* See {@link AMinimumDistanceSearchStrategy#useBiasCorrection}
*/
public ExhaustiveMinimumDistanceSearchStrategy(final boolean useBiasCorrection) {
super(useBiasCorrection);
}
/**
* Function returning the minimum distance among all subsequences of the given
* <code>timeSeries</code> to the <code>shapelet</code>'s data.
*
* @param shapelet
* The shapelet to be compared to all subsequences
* @param timeSeries
* The time series which subsequences are compared to the shapelet's
* data
* @return Return the minimum distance among all subsequences
*/
@Override
public double findMinimumDistance(final Shapelet shapelet, final double[] timeSeries) {
final int l = shapelet.getLength();
final int n = timeSeries.length;
double min = Double.MAX_VALUE;
double[] normalizedShapeletData = shapelet.getData();
// Reference implementation uses i < n-l => Leads sometimes to a better performance
for (int i = 0; i <= n - l; i++) {
double tmpED = MathUtil.singleSquaredEuclideanDistance(normalizedShapeletData, TimeSeriesUtil.zNormalize(TimeSeriesUtil.getInterval(timeSeries, i, i + l), this.useBiasCorrection));
if (tmpED < min) {
min = tmpED;
}
}
return min / l;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/shapelets
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/shapelets/search/package-info.java
|
/**
* This package contains search strategies applied to
* {@link ai.libs.jaicore.ml.classification.singlelabel.timeseries.shapelets.Shapelet} objects.
*
* @author Julian Lienen
*
*/
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.shapelets.search;
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/util/ClassMapper.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.util;
import java.util.List;
/**
* Class mapper used for predictions of String objects which are internally
* predicted by time series classifiers as ints.
*
* @author Julian Lienen
*
*/
public class ClassMapper {
/**
* Stored class values which indices are used to map integers to strings.
*/
private List<String> classValues;
/**
* Constructor using a list of String value to realize the mapping
*
* @param classValues
* String values used for the mapping. The values are identified by
* the given indices in the given list
*/
public ClassMapper(final List<String> classValues) {
this.classValues = classValues;
}
/**
* Maps a String value to an integer value based on the <code>value</code>'s
* position in the <code>classValues</code>.
*
* @param value
* The value to be looked up
* @return Returns the mapped index or -1 if not stored
*/
public int map(final String value) {
return this.classValues.indexOf(value);
}
/**
* Maps an integer value to a string based on the position <code>index</code> in
* the <code>classValues</code>.
*
* @param index
* The index used for the lookup
* @return Returns the given string at the position <code>index</code>
*/
public String map(final int index) {
return this.classValues.get(index);
}
/**
* Getter for the <code>classValues</code>.
*
* @return Returns the stored class values
*/
public List<String> getClassValues() {
return classValues;
}
/**
* Setter for the <code>classValues</code>.
*
* @param classValues
* The class values to be set.
*/
public void setClassValues(final List<String> classValues) {
this.classValues = classValues;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/util/HistogramBuilder.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.util;
import java.util.Arrays;
import java.util.HashMap;
import java.util.Map;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
/**
* @author Helen Beierling
* This class is used to compute Histograms for the found sfa words.
* This includes a numerosity reduction.
* (in form of double sequences which are used as key by using the Arrays class HashCode which are Integer).
* c.f. p. 1514 "The BOSS is concerned with time series classification in the presence of noise" by Patrick Schaefer
*/
public class HistogramBuilder {
private Map<Integer, Integer> histogram = new HashMap<>();
public Map<Integer, Integer> histogramForInstance(final TimeSeriesDataset2 blownUpSingleInstance) {
this.histogram.clear();
double[] lastWord = null;
// The blown up instance contains only one matrix.
for (double[] d : blownUpSingleInstance.getValues(0)) {
if (this.histogram.containsKey(Arrays.hashCode(d))) {
/*
* To the histogramm suczessiv duplicates are not added because of numerosity reduction.
* c.f.p.1514
* "The BOSS is concerned with time series classification in the presence of noise by Patrick Schaefer"
*/
if (!Arrays.equals(d, lastWord)) {
this.histogram.replace(Arrays.hashCode(d), this.histogram.get(Arrays.hashCode(d)) + 1);
}
} else {
this.histogram.put(Arrays.hashCode(d), 1);
}
lastWord = d;
}
return this.histogram;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/util/MathUtil.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.util;
import java.util.stream.DoubleStream;
/**
* Utility class consisting of mathematical utility functions.
*
* @author Julian Lienen
*
*/
public class MathUtil {
private MathUtil() {
/* no instantiation desired */
}
/**
* Function to calculate the sigmoid for the given value <code>z</code>.
*
* @param z
* Parameter z
* @return Returns the sigmoid for the parameter <code>z</code>.
*/
public static double sigmoid(final double z) {
return 1 / (1 + Math.exp((-1) * z));
}
/**
* Sums the values of the given <code>array</code>.
*
* @param array
* The array to be summed
* @return Returns the sum of the values
*/
public static double sum(final double[] array) {
return DoubleStream.of(array).sum();
}
/**
* Computes the single squared Euclidean distance between two vectors.
*
* @param vector1
* First argument vector
* @param vector2
* Second argument vector
* @return Returns the single squared Euclidean distance between two vectors
*/
public static double singleSquaredEuclideanDistance(final double[] vector1, final double[] vector2) {
if (vector1.length != vector2.length) {
throw new IllegalArgumentException("The lengths of of both vectors must match!");
}
double distance = 0;
for (int i = 0; i < vector1.length; i++) {
distance += Math.pow(vector1[i] - vector2[i], 2);
}
return distance;
}
/**
* Simple Manhattan distance (sum of the absolute differences between the
* vectors' elements) implementation for arrays of Integer.
*
* @param a
* First argument vector
* @param b
* Second argument vector
* @return Returns the Manhattan distance of the two given vectors
*/
public static double intManhattanDistance(final int[] a, final int[] b) {
double result = 0;
for (int j = 0; j < a.length; j++) {
result += Math.abs(a[j] - b[j]);
}
return result;
}
/**
* Function calculating the mean of the interval [t1, t2 (inclusive)] of the
* given <code>vector</code>.
*
* @param vector
* Vector which is used for the calculation
* @param t1
* Interval start
* @param t2
* Interval end (inclusive)
* @return Returns the mean of the vector's interval [t1, t2 (inclusive)]
*/
public static double mean(final double[] vector, final int t1, final int t2) {
checkIntervalParameters(vector, t1, t2);
double result = 0;
for (int i = t1; i <= t2; i++) {
result += vector[i];
}
return result / (t2 - t1 + 1);
}
/**
* Function calculating the standard deviation of the interval [t1, t2
* (inclusive)] of the given <code>vector</code>.
*
* @param vector
* Vector which is used for the calculation
* @param t1
* Interval start
* @param t2
* Interval end (inclusive)
* @param useBiasCorrection
* Indicator whether the bias (Bessel's) correction should be used
* @return Returns the standard deviation of the vector's interval [t1, t2
* (inclusive)]
*/
public static double stddev(final double[] vector, final int t1, final int t2, final boolean useBiasCorrection) {
checkIntervalParameters(vector, t1, t2);
if (t1 == t2) {
return 0.0d;
}
double mean = mean(vector, t1, t2);
double result = 0;
for (int i = t1; i <= t2; i++) {
result += Math.pow(vector[i] - mean, 2);
}
return Math.sqrt(result / (t2 - t1 + (useBiasCorrection ? 0 : 1)));
}
/**
* Function calculating the slope of the interval [t1, t2 (inclusive)] of the
* given <code>vector</code>.
*
* @param vector
* Vector which is used for the calculation
* @param t1
* Interval start
* @param t2
* Interval end (inclusive)
* @return Returns the slope of the vector's interval [t1, t2 (inclusive)]
*/
public static double slope(final double[] vector, final int t1, final int t2) {
checkIntervalParameters(vector, t1, t2);
if (t2 == t1) {
return 0d;
}
double xx = 0;
double x = 0;
double xy = 0;
double y = 0;
for (int i = t1; i <= t2; i++) {
x += i;
y += vector[i];
xx += i * i;
xy += i * vector[i];
}
// Calculate slope
int length = t2 - t1 + 1;
double denominator = (length * xx - x * x);
if (denominator == 0) {
throw new IllegalArgumentException("Given arguments yield a 0 " + denominator);
}
return (length * xy - x * y) / denominator;
}
/**
* Checks the parameters <code>t1</code> and </code>t2</code> for validity given
* the <code>vector</code>
*
* @param vector
* Value vector
* @param t1
* Interval start
* @param t2
* Interval end (inclusive)
*/
private static void checkIntervalParameters(final double[] vector, final int t1, final int t2) {
if (t1 >= vector.length || t2 >= vector.length) {
throw new IllegalArgumentException("Parameters t1 and t2 must be valid indices of the vector!");
}
if (t2 < t1) {
throw new IllegalArgumentException("End index t2 of the interval must be greater equals start index t1!");
}
}
/**
* Calculates the index of the maximum value in the given <code>array</code>
* (argmax).
*
* @param array
* Array to be checked. Must not be null or empty
* @return Returns the index of the maximum value
*/
public static int argmax(final int[] array) {
if (array == null || array.length == 0) {
throw new IllegalArgumentException("Given parameter 'array' must not be null or empty for argmax.");
}
int maxValue = array[0];
int index = 0;
for (int i = 1; i < array.length; i++) {
if (array[i] > maxValue) {
maxValue = array[i];
index = i;
}
}
return index;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/util/SimplifiedTimeSeriesLoader.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.util;
import java.io.BufferedInputStream;
import java.io.BufferedReader;
import java.io.File;
import java.io.FileInputStream;
import java.io.FileNotFoundException;
import java.io.IOException;
import java.io.InputStream;
import java.io.InputStreamReader;
import java.io.UnsupportedEncodingException;
import java.nio.charset.StandardCharsets;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.List;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import ai.libs.jaicore.basic.sets.Pair;
import ai.libs.jaicore.basic.sets.SetUtil;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.exception.TimeSeriesLoadingException;
/**
* Time series loader class which provides functionality to read datasets from
* files storing into simplified, more efficient time series datasets.
*
* @author Julian Lienen
*
*/
public class SimplifiedTimeSeriesLoader {
private SimplifiedTimeSeriesLoader() {
/* avoid instantiation */
}
/**
* Log4j logger.
*/
private static final Logger LOGGER = LoggerFactory.getLogger(SimplifiedTimeSeriesLoader.class);
/**
* Default charset used when extracting from files.
*/
public static final String DEFAULT_CHARSET = "UTF-8";
/**
* Prefix indicating an attribute declaration row in arff files.
*/
private static final String ARFF_ATTRIBUTE_PREFIX = "@attribute";
/**
* Delimiter in value enumerations in arff files.
*/
private static final String ARFF_VALUE_DELIMITER = ",";
/**
* Flag indicating the start of the data block in arff files.
*/
private static final String ARFF_DATA_FLAG = "@data";
/**
* Loads a univariate time series dataset from the given arff file. Assumes the
* class attribute to be the last among the declared attributes in the file.
*
* @param arffFile
* The arff file which is read
* @return Returns a pair consisting of an univariate TimeSeriesDataset object
* and a list of String objects containing the class values.
* @throws TimeSeriesLoadingException
* Throws an exception when the TimeSeriesDataset could not be
* created from the given file.
*/
@SuppressWarnings("unchecked")
public static Pair<TimeSeriesDataset2, ClassMapper> loadArff(final File arffFile) throws TimeSeriesLoadingException {
if (arffFile == null) {
throw new IllegalArgumentException("Parameter 'arffFile' must not be null!");
}
Object[] tsTargetClassNames = loadTimeSeriesWithTargetFromArffFile(arffFile);
ArrayList<double[][]> matrices = new ArrayList<>();
matrices.add((double[][]) tsTargetClassNames[0]);
ClassMapper cm = null;
if (tsTargetClassNames[2] != null) {
cm = new ClassMapper((List<String>) tsTargetClassNames[2]);
}
return new Pair<>(new TimeSeriesDataset2(matrices, new ArrayList<>(), (int[]) tsTargetClassNames[1]), cm);
}
/**
* Loads a multivariate time series dataset from multiple arff files (each for
* one series). The arff files must share the same targets among all series.
* Assumes the class attribute to be the last among the declared attributes in
* the file.
*
* @param arffFiles
* A sequence of arff files each containing one time series per
* instance
* @return Returns a multivariate TimeSeriesDataset object
* @throws TimeSeriesLoadingException
* Throws an exception when the TimeSeriesDataset could not be
* created from the given files.
*/
@SuppressWarnings("unchecked")
public static Pair<TimeSeriesDataset2, ClassMapper> loadArffs(final File... arffFiles) throws TimeSeriesLoadingException {
if (arffFiles == null) {
throw new IllegalArgumentException("Parameter 'arffFiles' must not be null!");
}
final List<double[][]> matrices = new ArrayList<>();
int[] target = null;
List<String> classNames = null;
for (final File arffFile : arffFiles) {
Object[] tsTargetClassNames = loadTimeSeriesWithTargetFromArffFile(arffFile);
if (classNames == null && tsTargetClassNames[2] != null) {
classNames = (List<String>) tsTargetClassNames[2];
} else {
// Check whether the same class names are used among all of the time series
List<String> furtherClassNames = (List<String>) tsTargetClassNames[2];
if ((classNames != null && furtherClassNames == null) || (furtherClassNames != null && !furtherClassNames.equals(classNames))) {
throw new TimeSeriesLoadingException("Could not load multivariate time series with different targets. Target values have to be stored in each " + "time series arff file and must be equal!");
}
}
if (target == null) {
target = (int[]) tsTargetClassNames[1];
} else {
// Check whether the same targets are used among all of the time series
int[] furtherTarget = (int[]) tsTargetClassNames[1];
if (furtherTarget == null || target.length != furtherTarget.length || !Arrays.equals(target, furtherTarget)) {
throw new TimeSeriesLoadingException("Could not load multivariate time series with different targets. Target values have to be stored in each " + "time series arff file and must be equal!");
}
}
// Check for same instance length
if (!matrices.isEmpty() && ((double[][]) tsTargetClassNames[0]).length != matrices.get(0).length) {
throw new TimeSeriesLoadingException("All time series must have the same first dimensionality (number of instances).");
}
matrices.add((double[][]) tsTargetClassNames[0]);
}
ClassMapper cm = null;
if (classNames != null) {
cm = new ClassMapper(classNames);
}
return new Pair<>(new TimeSeriesDataset2(matrices, new ArrayList<>(), target), cm);
}
/**
* Extracting the time series and target matrices from a given arff file.
* Assumes the class attribute to be the last among the declared attributes in
* the file.
*
* @param arffFile
* The arff file to be parsed
* @return Returns an object consisting of three elements: 1. The time series
* value matrix (double[][]), 2. the target value matrix (int[]) and 3.
* a list of the class value strings (List<String>)
* @throws TimeSeriesLoadingException
* Throws an exception when the matrices could not be extracted from
* the given arff file
*/
private static Object[] loadTimeSeriesWithTargetFromArffFile(final File arffFile) throws TimeSeriesLoadingException {
double[][] matrix = null;
int[] targetMatrix = null;
int numEmptyDataRows = 0;
List<String> targetValues = null;
boolean stringAttributes = false;
try (BufferedReader br = new BufferedReader(new InputStreamReader(new FileInputStream(arffFile), StandardCharsets.UTF_8))) {
int attributeCount = 0;
int lineCounter = 0;
int numInstances = 0;
final int fileLinesCount = countFileLines(arffFile);
boolean targetSet = false;
boolean readData = false;
String line;
String lastLine = "";
while ((line = br.readLine()) != null) {
if (!readData) {
lineCounter++;
// Set target values
if (!targetSet && line.equals("") && lastLine.startsWith(ARFF_ATTRIBUTE_PREFIX)) {
String targetString = lastLine.substring(lastLine.indexOf('{') + 1, lastLine.length() - 1);
targetValues = Arrays.asList(targetString.split(ARFF_VALUE_DELIMITER));
if (!SetUtil.doesStringCollectionOnlyContainNumbers(targetValues)) {
stringAttributes = true;
}
targetSet = true;
}
// Count attributes
if (line.startsWith(ARFF_ATTRIBUTE_PREFIX)) {
attributeCount++;
}
if (line.startsWith(ARFF_DATA_FLAG)) {
readData = true;
numInstances = fileLinesCount - lineCounter + 1;
matrix = new double[numInstances][targetSet ? attributeCount - 1 : attributeCount];
targetMatrix = new int[numInstances];
lineCounter = 0;
if (!targetSet) {
LOGGER.warn("No target has been set before reading data.");
}
}
} else {
if (!line.equals("")) {
// Read the data
String[] values = line.split(ARFF_VALUE_DELIMITER);
double[] dValues = new double[targetSet ? values.length - 1 : values.length];
for (int i = 0; i < values.length - 1; i++) {
dValues[i] = Double.parseDouble(values[i]);
}
matrix[lineCounter] = dValues;
if (targetSet) {
targetMatrix[lineCounter] = targetValues.indexOf(values[values.length - 1]);
}
}
lineCounter++;
}
lastLine = line;
}
// Update empty data rows
numEmptyDataRows = numInstances - lineCounter;
if (matrix == null) {
throw new IllegalStateException("Matrix is null, which it should not be at this point!");
}
// Due to efficiency reasons, the matrices are narrowed afterwards to eliminate
// empty data rows
if (numEmptyDataRows > 0) {
int endIndex = matrix.length - numEmptyDataRows;
matrix = getInterval(matrix, 0, endIndex);
targetMatrix = getInterval(targetMatrix, 0, endIndex);
}
Object[] result = new Object[3];
result[0] = matrix;
result[1] = targetMatrix;
result[2] = stringAttributes ? targetValues : null;
return result;
} catch (UnsupportedEncodingException e) {
throw new TimeSeriesLoadingException("Could not load time series dataset due to unsupported encoding.", e);
} catch (FileNotFoundException e) {
throw new TimeSeriesLoadingException(String.format("Could not locate time series dataset file '%s'.", arffFile.getPath()), e);
} catch (IOException e) {
throw new TimeSeriesLoadingException("Could not load time series dataset due to IOException.", e);
}
}
/**
* Function returning a submatrix of the given <code>matrix</code>. The
* submatrix is specified by the indices <code>begin</code> and
* </code>end</code> (exclusive). Only the rows within this interval are copied
* into the result matrix.
*
* @param matrix
* The matrix from which the submatrix is extracted
* @param begin
* Begin index of the rows to be extracted
* @param end
* Exclusive end index of the rows to be extracted
* @return Returns the specified submatrix
*/
private static double[][] getInterval(final double[][] matrix, final int begin, final int end) {
if (begin < 0 || begin > matrix.length - 1) {
throw new IllegalArgumentException("The begin index must be valid!");
}
if (end < 1 || end > matrix.length) {
throw new IllegalArgumentException("The end index must be valid!");
}
final double[][] result = new double[end - begin][];
for (int i = 0; i < end - begin; i++) {
result[i] = matrix[i + begin];
}
return result;
}
/**
* Function returning an interval as subarray of the given <code>array</code>.
* The interval is specified by the indices <code>begin</code> and
* </code>end</code> (exclusive).
*
* @param array
* The array from which the interval is extracted
* @param begin
* Begin index of the interval
* @param end
* Exclusive end index of the interval
* @return Returns the specified interval as a subarray
*/
private static int[] getInterval(final int[] array, final int begin, final int end) {
if (begin < 0 || begin > array.length - 1) {
throw new IllegalArgumentException("The begin index must be valid!");
}
if (end < 1 || end > array.length) {
throw new IllegalArgumentException("The end index must be valid!");
}
final int[] result = new int[end - begin];
for (int i = 0; i < end - begin; i++) {
result[i] = array[i + begin];
}
return result;
}
/**
* Counts the lines of the given File object in a very efficient way (thanks to
* https://stackoverflow.com/a/453067).
*
* @param filename
* File which lines of code are counted
* @return Returns the number of file lines
* @throws IOException
* Throws exception when the given file could not be read
*/
public static int countFileLines(final File file) throws IOException {
try (InputStream is = new BufferedInputStream(new FileInputStream(file))) {
byte[] c = new byte[1024];
int readChars = is.read(c);
if (readChars == -1) {
// bail out if nothing to read
return 0;
}
// make it easy for the optimizer to tune this loop
int count = 0;
while (readChars == 1024) {
for (int i = 0; i < 1024; i++) {
if (c[i] == '\n') {
++count;
}
}
readChars = is.read(c);
}
// count remaining characters
while (readChars != -1) {
for (int i = 0; i < readChars; ++i) {
if (c[i] == '\n') {
++count;
}
}
readChars = is.read(c);
}
return count == 0 ? 1 : count;
}
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/util/TSLearningProblem.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.util;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset2;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.quality.IQualityMeasure;
public class TSLearningProblem {
private final IQualityMeasure qualityMeasure;
private final TimeSeriesDataset2 dataset;
public TSLearningProblem(final IQualityMeasure qualityMeasure, final TimeSeriesDataset2 dataset) {
super();
this.qualityMeasure = qualityMeasure;
this.dataset = dataset;
}
public IQualityMeasure getQualityMeasure() {
return this.qualityMeasure;
}
public TimeSeriesDataset2 getDataset() {
return this.dataset;
}
}
|
0
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries
|
java-sources/ai/libs/jaicore-ml/0.2.7/ai/libs/jaicore/ml/classification/singlelabel/timeseries/util/TimeSeriesBatchLoader.java
|
package ai.libs.jaicore.ml.classification.singlelabel.timeseries.util;
import ai.libs.jaicore.ml.classification.singlelabel.timeseries.dataset.TimeSeriesDataset;
/**
* BatchLoader
*/
public class TimeSeriesBatchLoader {
public TimeSeriesBatchLoader(TimeSeriesDataset dataset, int batchSize, boolean shuffle) {
}
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.