Examples with ClassifierLearner edu.cmu.minorthird.classify.ClassifierLearner used on opensource projects

Example 1 with ClassifierLearner

use of edu.cmu.minorthird.classify.ClassifierLearner in project lucida by claritylab.

the class HierarchicalClassifierLearner method addExample.

public void addExample(Example example) {
    for (int i = 0; i < prototypes.length; i++) {
        String labelName = example.getLabel().bestClassName();
        String prefix = getLabelPrefix(labelName, i);
        String sublabel = getSublabel(labelName, i);
        Example subExample = new Example(example.asInstance(), new ClassLabel(sublabel));
        ClassifierLearner subLearner = classifierLearners.get(prefix);
        subLearner.addExample(subExample);
    }
}

Example 2 with ClassifierLearner

use of edu.cmu.minorthird.classify.ClassifierLearner in project lucida by claritylab.

the class HierarchicalClassifierLearner method setSchema.

public void setSchema(ExampleSchema schema) {
    String[] labelNames = schema.validClassNames();
    for (int i = 0; i < labelNames.length; i++) {
        for (int j = 0; j < prototypes.length; j++) {
            String prefix = getLabelPrefix(labelNames[i], j);
            if (!classifierLearners.containsKey(prefix)) {
                System.out.println("Making new schema and learner for " + prefix);
                ExampleSchema subSchema = createSubSchema(schema, prefix, j);
                ClassifierLearner newLearner;
                if (subSchema.getNumberOfClasses() == 1) {
                    System.out.println("Only 1 class to learn for " + prefix + "; using DummyClassifier and Learner");
                    newLearner = new DummyClassifierLearner(subSchema.getClassName(0));
                } else {
                    newLearner = prototypes[j].copy();
                    newLearner.setSchema(subSchema);
                }
                classifierLearners.put(prefix, newLearner);
            }
        }
    }
}

Example 3 with ClassifierLearner

use of edu.cmu.minorthird.classify.ClassifierLearner in project lucida by claritylab.

the class HierarchicalClassifierTrainer method createLearnerByName.

public ClassifierLearner createLearnerByName(String name) {
    ClassifierLearner learner;
    //K-Nearest-Neighbor learner, using m3rd recommended parameters
    if (name.equalsIgnoreCase("KNN")) {
        learner = new KnnLearner();
    } else //K-Way Mixture learner, using m3rd recommended parameters
    if (name.equalsIgnoreCase("KWAY_MIX")) {
        learner = new KWayMixtureLearner();
    } else //Maximum Entropy learner, using m3rd recommended parameters
    if (name.equalsIgnoreCase("MAX_ENT")) {
        learner = new MaxEntLearner();
    } else //Balanced Winnow learner with One vs All binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("BWINNOW_OVA")) {
        learner = new OneVsAllLearner(new BalancedWinnow());
    } else //Margin Perceptron learner with One vs All binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("MPERCEPTRON_OVA")) {
        learner = new OneVsAllLearner(new MarginPerceptron());
    } else //Naive Bayes learner with One vs All binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("NBAYES_OVA")) {
        learner = new OneVsAllLearner(new NaiveBayes());
    } else //Voted Perceptron learner with One vs All binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("VPERCEPTRON_OVA")) {
        learner = new OneVsAllLearner(new VotedPerceptron());
    } else //Ada Boost learner with One vs All binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("ADABOOST_OVA")) {
        learner = new OneVsAllLearner(new AdaBoost());
    } else //Ada Boost learner with Cascading binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("ADABOOST_CB")) {
        learner = new CascadingBinaryLearner(new AdaBoost());
    } else //Ada Boost learner with Most Frequent First binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("ADABOOST_MFF")) {
        learner = new MostFrequentFirstLearner(new AdaBoost());
    } else //Ada Boost learner (Logistic Regression version) with One vs All binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("ADABOOSTL_OVA")) {
        learner = new OneVsAllLearner(new AdaBoost.L());
    } else //Ada Boost learner (Logistic Regression version) with Cascading binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("ADABOOSTL_CB")) {
        learner = new CascadingBinaryLearner(new AdaBoost.L());
    } else //Ada Boost learner (Logistic Regression version) with Most Frequent First binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("ADABOOSTL_MFF")) {
        learner = new MostFrequentFirstLearner(new AdaBoost.L());
    } else //Decision Tree learner with One vs All binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("DTREE_OVA")) {
        learner = new OneVsAllLearner(new DecisionTreeLearner());
    } else //Decision Tree learner with Cascading binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("DTREE_CB")) {
        learner = new CascadingBinaryLearner(new DecisionTreeLearner());
    } else //Decision Tree learner with Most Frequent First binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("DTREE_MFF")) {
        learner = new MostFrequentFirstLearner(new DecisionTreeLearner());
    } else //Negative Binomial learner with One vs All binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("NEGBI_OVA")) {
        learner = new OneVsAllLearner(new NegativeBinomialLearner());
    } else //Negative Binomial learner with Cascading binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("NEGBI_CB")) {
        learner = new CascadingBinaryLearner(new NegativeBinomialLearner());
    } else //Negative Binomial learner with Most Frequent First binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("NEGBI_MFF")) {
        learner = new MostFrequentFirstLearner(new NegativeBinomialLearner());
    } else //SVM learner with One vs All binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("SVM_OVA")) {
        learner = new OneVsAllLearner(new SVMLearner());
    } else //SVM learner with One vs All binary transformer, using testing parameters
    if (name.equalsIgnoreCase("SVM_OVA_CONF1")) {
        svm_parameter param = new svm_parameter();
        param.svm_type = svm_parameter.C_SVC;
        param.kernel_type = svm_parameter.POLY;
        param.degree = 2;
        // 1/k
        param.gamma = 1;
        param.coef0 = 0;
        param.nu = 0.5;
        param.cache_size = 40;
        param.C = 1;
        param.eps = 1e-3;
        param.p = 0.1;
        param.shrinking = 1;
        param.nr_weight = 0;
        param.weight_label = new int[0];
        param.weight = new double[0];
        learner = new OneVsAllLearner(new SVMLearner(param));
    } else //SVM learner with Cascading binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("SVM_CB")) {
        learner = new CascadingBinaryLearner(new SVMLearner());
    } else //SVM learner with Most Frequent First binary transformer, using m3rd recommended parameters
    if (name.equalsIgnoreCase("SVM_MFF")) {
        learner = new MostFrequentFirstLearner(new SVMLearner());
    } else {
        System.err.println("Unrecognized learner name: " + name);
        learner = null;
    }
    return learner;
}

Example 4 with ClassifierLearner

use of edu.cmu.minorthird.classify.ClassifierLearner in project lucida by claritylab.

the class ScoreNormalizationFilter method evaluate.

/**
	 * Performs a cross-validation on the given data set for the given features
	 * and model.
	 * 
	 * @param serializedDir directory containing serialized results
	 * @param features selected features
	 * @param model selected model
	 * @return evaluation statistics
	 */
public static Evaluation evaluate(String serializedDir, String[] features, String model) {
    // create data set with selected features from serialized results
    Dataset dataSet = createDataset(features, serializedDir);
    // create learner for selected model
    ClassifierLearner learner = createLearner(model);
    // cross-validate model on data set
    RandomElement r = new RandomElement(System.currentTimeMillis());
    Splitter splitter = new CrossValSplitter(r, NUM_FOLDS);
    CrossValidatedDataset cvDataset = new CrossValidatedDataset(learner, dataSet, splitter, true);
    Evaluation eval = cvDataset.getEvaluation();
    return eval;
}

Example 5 with ClassifierLearner

use of edu.cmu.minorthird.classify.ClassifierLearner in project lucida by claritylab.

the class ScoreNormalizationFilter method train.

/**
	 * Trains a classifier using the given training data, features and model.
	 * 
	 * @param serializedDir directory containing serialized results
	 * @param features selected features
	 * @param model selected model
	 * @return trained classifier
	 */
public static Classifier train(String serializedDir, String[] features, String model) {
    // create training set with given features from serialized results
    Dataset trainingSet = createDataset(features, serializedDir);
    // create learner for given model
    ClassifierLearner learner = createLearner(model);
    // train classifier
    Classifier classifier = new DatasetClassifierTeacher(trainingSet).train(learner);
    return classifier;
}