@Override
public Model learn(ExampleSet exampleSet) throws OperatorException {
Kernel kernel = getKernel();
kernel.init(exampleSet);
double initLearnRate = getParameterAsDouble(PARAMETER_LEARNING_RATE);
NominalMapping labelMapping = exampleSet.getAttributes().getLabel().getMapping();
String classNeg = labelMapping.getNegativeString();
String classPos = labelMapping.getPositiveString();
double classValueNeg = labelMapping.getNegativeIndex();
int numberOfAttributes = exampleSet.getAttributes().size();
HyperplaneModel model = new HyperplaneModel(exampleSet, classNeg, classPos, kernel);
model.init(new double[numberOfAttributes], 0);
for (int round = 0; round <= getParameterAsInt(PARAMETER_ROUNDS); round++) {
double learnRate = getLearnRate(round, getParameterAsInt(PARAMETER_ROUNDS), initLearnRate);
Attributes attributes = exampleSet.getAttributes();
for (Example example : exampleSet) {
double prediction = model.predict(example);
if (prediction != example.getLabel()) {
double direction = (example.getLabel() == classValueNeg) ? -1 : 1;
// adapting intercept
model.setIntercept(model.getIntercept() + learnRate * direction);
// adapting coefficients
double coefficients[] = model.getCoefficients();
int i = 0;
for (Attribute attribute : attributes) {
coefficients[i] += learnRate * direction * example.getValue(attribute);
i++;
}
}
}
}
return model;
}
@Override
public void doWork() throws OperatorException {
ExampleSet exampleSet = exampleSetInput.getData(ExampleSet.class);
// only use numeric attributes
Tools.onlyNumericalAttributes(exampleSet, "KernelPCA");
Tools.onlyNonMissingValues(exampleSet, getOperatorClassName(), this);
Attributes attributes = exampleSet.getAttributes();
int numberOfExamples = exampleSet.size();
// calculating means for later zero centering
exampleSet.recalculateAllAttributeStatistics();
double[] means = new double[exampleSet.getAttributes().size()];
int i = 0;
for (Attribute attribute : exampleSet.getAttributes()) {
means[i] = exampleSet.getStatistics(attribute, Statistics.AVERAGE);
i++;
}
// kernel
Kernel kernel = Kernel.createKernel(this);
// copying zero centered exampleValues
ArrayList<double[]> exampleValues = new ArrayList<double[]>(numberOfExamples);
i = 0;
for (Example columnExample : exampleSet) {
double[] columnValues = getAttributeValues(columnExample, attributes, means);
exampleValues.add(columnValues);
i++;
}
// filling kernel matrix
Matrix kernelMatrix = new Matrix(numberOfExamples, numberOfExamples);
for (i = 0; i < numberOfExamples; i++) {
for (int j = 0; j < numberOfExamples; j++) {
kernelMatrix.set(
i, j, kernel.calculateDistance(exampleValues.get(i), exampleValues.get(j)));
}
}
// calculating eigenVectors
EigenvalueDecomposition eig = kernelMatrix.eig();
Model model = new KernelPCAModel(exampleSet, means, eig.getV(), exampleValues, kernel);
if (exampleSetOutput.isConnected()) {
exampleSetOutput.deliver(model.apply(exampleSet));
}
originalOutput.deliver(exampleSet);
modelOutput.deliver(model);
}
@Override
public List<ParameterType> getParameterTypes() {
List<ParameterType> types = super.getParameterTypes();
types.add(
new ParameterTypeBoolean(
PARAMETER_USE_WEIGHTS,
"Indicates if the weight attribute should be used.",
false,
false));
types.add(
new ParameterTypeInt(
PARAMETER_K,
"The number of clusters which should be detected.",
2,
Integer.MAX_VALUE,
2,
false));
types.add(
new ParameterTypeInt(
PARAMETER_MAX_OPTIMIZATION_STEPS,
"The maximal number of iterations performed for one run of k-Means.",
1,
Integer.MAX_VALUE,
100,
false));
types.addAll(RandomGenerator.getRandomGeneratorParameters(this));
types.addAll(Kernel.getParameters(this));
return types;
}
@Override
public double predict(Example example) throws OperatorException {
int i = 0;
double distance = intercept;
// using kernel for distance calculation
double[] values = new double[example.getAttributes().size()];
for (Attribute currentAttribute : example.getAttributes()) {
values[i] = example.getValue(currentAttribute);
i++;
}
distance += kernel.calculateDistance(values, coefficients);
if (getLabel().isNominal()) {
int positiveMapping = getLabel().getMapping().mapString(classPositive);
int negativeMapping = getLabel().getMapping().mapString(classNegative);
boolean isApplying = example.getAttributes().getPredictedLabel() != null;
if (isApplying) {
example.setConfidence(classPositive, 1.0d / (1.0d + java.lang.Math.exp(-distance)));
example.setConfidence(classNegative, 1.0d / (1.0d + java.lang.Math.exp(distance)));
}
if (distance < 0) {
return negativeMapping;
} else {
return positiveMapping;
}
} else {
return distance;
}
}
@Override
public ClusterModel generateClusterModel(ExampleSet exampleSet) throws OperatorException {
int k = getParameterAsInt(PARAMETER_K);
int maxOptimizationSteps = getParameterAsInt(PARAMETER_MAX_OPTIMIZATION_STEPS);
boolean useExampleWeights = getParameterAsBoolean(PARAMETER_USE_WEIGHTS);
Kernel kernel = Kernel.createKernel(this);
// init operator progress
getProgress().setTotal(maxOptimizationSteps);
// checking and creating ids if necessary
Tools.checkAndCreateIds(exampleSet);
// additional checks
Tools.onlyNonMissingValues(exampleSet, getOperatorClassName(), this, new String[0]);
if (exampleSet.size() < k) {
throw new UserError(this, 142, k);
}
// extracting attribute names
Attributes attributes = exampleSet.getAttributes();
ArrayList<String> attributeNames = new ArrayList<String>(attributes.size());
for (Attribute attribute : attributes) {
attributeNames.add(attribute.getName());
}
Attribute weightAttribute = attributes.getWeight();
RandomGenerator generator = RandomGenerator.getRandomGenerator(this);
ClusterModel model =
new ClusterModel(
exampleSet,
k,
getParameterAsBoolean(RMAbstractClusterer.PARAMETER_ADD_AS_LABEL),
getParameterAsBoolean(RMAbstractClusterer.PARAMETER_REMOVE_UNLABELED));
// init centroids
int[] clusterAssignments = new int[exampleSet.size()];
for (int i = 0; i < exampleSet.size(); i++) {
clusterAssignments[i] = generator.nextIntInRange(0, k);
}
// run optimization steps
boolean stable = false;
for (int step = 0; step < maxOptimizationSteps && !stable; step++) {
// calculating cluster kernel properties
double[] clusterWeights = new double[k];
double[] clusterKernelCorrection = new double[k];
int i = 0;
for (Example firstExample : exampleSet) {
double firstExampleWeight = useExampleWeights ? firstExample.getValue(weightAttribute) : 1d;
double[] firstExampleValues = getAsDoubleArray(firstExample, attributes);
clusterWeights[clusterAssignments[i]] += firstExampleWeight;
int j = 0;
for (Example secondExample : exampleSet) {
if (clusterAssignments[i] == clusterAssignments[j]) {
double secondExampleWeight =
useExampleWeights ? secondExample.getValue(weightAttribute) : 1d;
clusterKernelCorrection[clusterAssignments[i]] +=
firstExampleWeight
* secondExampleWeight
* kernel.calculateDistance(
firstExampleValues, getAsDoubleArray(secondExample, attributes));
}
j++;
}
i++;
}
for (int z = 0; z < k; z++) {
clusterKernelCorrection[z] /= clusterWeights[z] * clusterWeights[z];
}
// assign examples to new centroids
int[] newClusterAssignments = new int[exampleSet.size()];
i = 0;
for (Example example : exampleSet) {
double[] exampleValues = getAsDoubleArray(example, attributes);
double exampleKernelValue = kernel.calculateDistance(exampleValues, exampleValues);
double nearestDistance = Double.POSITIVE_INFINITY;
int nearestIndex = 0;
for (int clusterIndex = 0; clusterIndex < k; clusterIndex++) {
double distance = 0;
// iterating over all examples in cluster to get kernel distance
int j = 0;
for (Example clusterExample : exampleSet) {
if (clusterAssignments[j] == clusterIndex) {
distance +=
(useExampleWeights ? clusterExample.getValue(weightAttribute) : 1d)
* kernel.calculateDistance(
getAsDoubleArray(clusterExample, attributes), exampleValues);
}
j++;
}
distance *= -2d / clusterWeights[clusterIndex];
// copy in outer loop
distance += exampleKernelValue;
distance += clusterKernelCorrection[clusterIndex];
if (distance < nearestDistance) {
nearestDistance = distance;
nearestIndex = clusterIndex;
}
}
newClusterAssignments[i] = nearestIndex;
i++;
}
// finishing assignment
stable = true;
for (int j = 0; j < exampleSet.size() && stable; j++) {
stable &= newClusterAssignments[j] == clusterAssignments[j];
}
clusterAssignments = newClusterAssignments;
// trigger operator progress
getProgress().step();
}
// setting last clustering into model
model.setClusterAssignments(clusterAssignments, exampleSet);
getProgress().complete();
if (addsClusterAttribute()) {
Attribute cluster = AttributeFactory.createAttribute("cluster", Ontology.NOMINAL);
exampleSet.getExampleTable().addAttribute(cluster);
exampleSet.getAttributes().setCluster(cluster);
int i = 0;
for (Example example : exampleSet) {
example.setValue(cluster, "cluster_" + clusterAssignments[i]);
i++;
}
}
return model;
}
@Override
public List<ParameterType> getParameterTypes() {
List<ParameterType> types = super.getParameterTypes();
types.addAll(Kernel.getParameters(this));
return types;
}
/** Returns a model containing all support vectors, i.e. the examples with non-zero alphas. */
private EvoSVMModel getModel(double[] alphas) {
// calculate support vectors
Iterator<Example> reader = exampleSet.iterator();
List<SupportVector> supportVectors = new ArrayList<SupportVector>();
int index = 0;
while (reader.hasNext()) {
double currentAlpha = alphas[index];
Example currentExample = reader.next();
if (currentAlpha != 0.0d) {
double[] x = new double[exampleSet.getAttributes().size()];
int a = 0;
for (Attribute attribute : exampleSet.getAttributes())
x[a++] = currentExample.getValue(attribute);
supportVectors.add(new SupportVector(x, ys[index], currentAlpha));
}
index++;
}
// calculate all sum values
double[] sum = new double[exampleSet.size()];
reader = exampleSet.iterator();
index = 0;
while (reader.hasNext()) {
Example current = reader.next();
double[] x = new double[exampleSet.getAttributes().size()];
int a = 0;
for (Attribute attribute : exampleSet.getAttributes()) x[a++] = current.getValue(attribute);
sum[index] = kernel.getSum(supportVectors, x);
index++;
}
// calculate b (from Stefan's mySVM code)
double bSum = 0.0d;
int bCounter = 0;
for (int i = 0; i < alphas.length; i++) {
if ((ys[i] * alphas[i] - c < -IS_ZERO) && (ys[i] * alphas[i] > IS_ZERO)) {
bSum += ys[i] - sum[i];
bCounter++;
} else if ((ys[i] * alphas[i] + c > IS_ZERO) && (ys[i] * alphas[i] < -IS_ZERO)) {
bSum += ys[i] - sum[i];
bCounter++;
}
}
if (bCounter == 0) {
// unlikely
bSum = 0.0d;
for (int i = 0; i < alphas.length; i++) {
if ((ys[i] * alphas[i] < IS_ZERO) && (ys[i] * alphas[i] > -IS_ZERO)) {
bSum += ys[i] - sum[i];
bCounter++;
}
}
if (bCounter == 0) {
// even unlikelier
bSum = 0.0d;
for (int i = 0; i < alphas.length; i++) {
bSum += ys[i] - sum[i];
bCounter++;
}
}
}
return new EvoSVMModel(exampleSet, supportVectors, kernel, bSum / bCounter);
}