protected static void predict(String[] args) {
if (args.length != 3) {
System.out.println(
"Please specify the path for the test data as the second argument "
+ "and the path to save the predictions as the third argument.");
} else {
final String testDataFilename = args[1];
try {
// Parse the sparse testing data.
final Map.Entry<List<SparseLineAST>, Integer> instancesToMaxFeatureIndex =
SparseLineAST.parseSparseEncodingFromInputFilename(
testDataFilename, new SparseLineAST.NumericValidator());
final List<SparseLineAST> instances = instancesToMaxFeatureIndex.getKey();
final int maxFeatureIndex = instancesToMaxFeatureIndex.getValue();
// Initialize the classifier from the file data.
DiscretizingNaiveBayesClassifier classifier = new DiscretizingNaiveBayesClassifier();
classifier.initializeFromFileData(loadTrainedModelFileData());
// Make predictions.
// int numCorrectPredictions = 0; // TODO: Remove debug code.
List<Integer> predictedClassIndices = new ArrayList<Integer>(instances.size());
for (SparseLineAST instance : instances) {
final int predictedClassIndex = classifier.getPrediction(instance);
predictedClassIndices.add(predictedClassIndex);
/*
// TODO: Remove debug code.
final int correctClassIndex = Integer.parseInt(instance.getOutput());
if (predictedClassIndex == correctClassIndex) {
numCorrectPredictions += 1;
}
//*/
}
/*
// TODO: Remove debug code.
System.out.println(String.format("%d of %d predicted correctly (%f).",
numCorrectPredictions, predictedClassIndices.size(),
numCorrectPredictions / (double) predictedClassIndices.size()));
//*/
// Save the predictions to file.
final String predictionsFilename = args[2];
savePredictions(predictionsFilename, predictedClassIndices);
} catch (RuntimeException x) {
System.out.println(x.getMessage());
}
}
}
protected static void generate(String[] args) {
if (args.length != 3) {
System.out.println(
"Please specify the number of instances as the second argument "
+ "and the path to save the instances as the third argument.");
} else {
final int numInstances = Integer.parseInt(args[1]);
final String dataFilename = args[2];
// Initialize the classifier from the file data.
DiscretizingNaiveBayesClassifier classifier = new DiscretizingNaiveBayesClassifier();
classifier.initializeFromFileData(loadTrainedModelFileData());
SparseLineAST.saveSparseEncoding(
dataFilename, classifier.generateLabeledInstances(numInstances));
}
}
protected static void randomLearn(String[] args) {
if (args.length != 3) {
System.out.println(
"Please specify the number of classes as the second argument "
+ "and the number of features as the third argument.");
} else {
final int numClassIndices = Integer.parseInt(args[1]);
final int maxFeatureIndex = Integer.parseInt(args[2]);
// Initialize the classifier to random parameters.
DiscretizingNaiveBayesClassifier classifier = new DiscretizingNaiveBayesClassifier();
classifier.initializeRandomly(
DiscretizingNaiveBayesClassifier.makeClassIndexSet(numClassIndices),
maxFeatureIndex,
new DiscretizingNaiveBayesClassifier.SimpleBinaryDiscretizer());
// Save the trained model.
saveTrainedModelFileData(classifier.toFileData());
}
}
protected static void semiSupervisedLearn(String[] args) {
if (args.length < 2) {
System.out.println(
"Please specify the path for the training data as the second argument. "
+ "(Optionally, you can give the number of EM iterations as the third argument.)");
} else {
final String trainingDataFilename = args[1];
int maxNumIterations;
if (args.length == 3) {
maxNumIterations = Integer.parseInt(args[2]);
} else {
maxNumIterations = Integer.MAX_VALUE;
}
try {
// Parse the sparse training data.
final Map.Entry<List<SparseLineAST>, Integer> instancesToMaxFeatureIndex =
SparseLineAST.parseSparseEncodingFromInputFilename(
trainingDataFilename,
new SparseLineAST.SemiSupervisedTrainingValidator(
new SparseLineAST.NumericValidator()));
final List<SparseLineAST> instances = instancesToMaxFeatureIndex.getKey();
final int maxFeatureIndex = instancesToMaxFeatureIndex.getValue();
// Initialize and train the classifier on the training data.
DiscretizingNaiveBayesClassifier classifier =
DiscretizingNaiveBayesClassifierEMTrainer.train(
getKForSmoothing().getSecond(),
instances,
maxFeatureIndex,
new DiscretizingNaiveBayesClassifier.SimpleBinaryDiscretizer(),
maxNumIterations);
// Save the trained model.
saveTrainedModelFileData(classifier.toFileData());
} catch (RuntimeException x) {
System.out.println(x.getMessage());
}
}
}
protected static void learn(String[] args) {
if (args.length != 2) {
System.out.println("Please specify the path for the training data as the second argument.");
} else {
final String trainingDataFilename = args[1];
try {
// Parse the sparse training data.
// Note: "u" outputs for training instances are accepted; they'll just be ignored.
final Map.Entry<List<SparseLineAST>, Integer> instancesToMaxFeatureIndex =
SparseLineAST.parseSparseEncodingFromInputFilename(
trainingDataFilename,
new SparseLineAST.SemiSupervisedTrainingValidator(
new SparseLineAST.NumericValidator()));
final List<SparseLineAST> instances = instancesToMaxFeatureIndex.getKey();
final int maxFeatureIndex = instancesToMaxFeatureIndex.getValue();
// Ignore unlabeled instances.
final List<SparseLineAST> labeledInstances = SparseLineAST.getLabeledInstances(instances);
// Initialize and train the classifier on the training data.
DiscretizingNaiveBayesClassifier classifier = new DiscretizingNaiveBayesClassifier();
classifier.initialize(getKForSmoothing().getSecond());
classifier.train(
labeledInstances,
maxFeatureIndex,
new DiscretizingNaiveBayesClassifier.SimpleBinaryDiscretizer());
// Save the trained model.
saveTrainedModelFileData(classifier.toFileData());
} catch (RuntimeException x) {
System.out.println(x.getMessage());
for (StackTraceElement element : x.getStackTrace()) {
System.out.println(element);
}
}
}
}
