Examples with LogisticRegression edu.neu.ccs.pyramid.classification.logistic_regression.LogisticRegression used on opensource projects

Example 11 with LogisticRegression

use of edu.neu.ccs.pyramid.classification.logistic_regression.LogisticRegression in project pyramid by cheng-li.

the class CBMInspector method distanceFromMean.

public static double distanceFromMean(CBM bmm, int label) {
    Classifier.ProbabilityEstimator[][] logistics = bmm.getBinaryClassifiers();
    int numClusters = bmm.getNumComponents();
    int numFeatures = ((LogisticRegression) logistics[0][0]).getNumFeatures();
    Vector positiveAverageVector = new DenseVector(numFeatures);
    for (int k = 0; k < numClusters; k++) {
        Vector positiveVector = ((LogisticRegression) logistics[k][label]).getWeights().getWeightsWithoutBiasForClass(1);
        positiveAverageVector = positiveAverageVector.plus(positiveVector);
    }
    positiveAverageVector = positiveAverageVector.divide(numClusters);
    double dis = 0;
    for (int k = 0; k < numClusters; k++) {
        Vector positiveVector = ((LogisticRegression) logistics[k][label]).getWeights().getWeightsWithoutBiasForClass(1);
        dis += positiveVector.minus(positiveAverageVector).norm(2);
    }
    return dis / numClusters;
}

Example 12 with LogisticRegression

use of edu.neu.ccs.pyramid.classification.logistic_regression.LogisticRegression in project pyramid by cheng-li.

the class MLPlattScaling method fitClassK.

private static LogisticRegression fitClassK(double[] scores, int[] labels) {
    ClfDataSet dataSet = ClfDataSetBuilder.getBuilder().numClasses(2).numDataPoints(scores.length).numFeatures(1).dense(true).missingValue(false).build();
    for (int i = 0; i < scores.length; i++) {
        dataSet.setFeatureValue(i, 0, scores[i]);
        dataSet.setLabel(i, labels[i]);
    }
    LogisticRegression logisticRegression = new LogisticRegression(2, dataSet.getNumFeatures());
    ElasticNetLogisticTrainer trainer = ElasticNetLogisticTrainer.newBuilder(logisticRegression, dataSet).setRegularization(1.0E-9).setL1Ratio(0).build();
    trainer.optimize();
    return logisticRegression;
}

Example 13 with LogisticRegression

use of edu.neu.ccs.pyramid.classification.logistic_regression.LogisticRegression in project pyramid by cheng-li.

the class CBMUtilityOptimizer method updateBinaryLogisticRegression.

private void updateBinaryLogisticRegression(int componentIndex, int labelIndex) {
    RidgeLogisticOptimizer ridgeLogisticOptimizer;
    //        System.out.println("for component "+componentIndex+"  label "+labelIndex);
    //        System.out.println("weights="+Arrays.toString(gammasT[componentIndex]));
    //        System.out.println("binary target distribution="+Arrays.deepToString(binaryTargetsDistributions[labelIndex]));
    //        double posProb = 0;
    //        double negProb = 0;
    //        for (int i=0;i<dataSet.getNumDataPoints();i++){
    //            posProb += gammasT[componentIndex][i] * binaryTargetsDistributions[labelIndex][i][1];
    //            negProb += gammasT[componentIndex][i] * binaryTargetsDistributions[labelIndex][i][0];
    //        }
    //        System.out.println("sum pos prob = "+posProb);
    //        System.out.println("sum neg prob = "+negProb);
    // no parallelism
    ridgeLogisticOptimizer = new RidgeLogisticOptimizer((LogisticRegression) cbm.binaryClassifiers[componentIndex][labelIndex], dataSet, gammasT[componentIndex], binaryTargetsDistributions[labelIndex], priorVarianceBinary, false);
    //TODO maximum iterations
    ridgeLogisticOptimizer.getOptimizer().getTerminator().setMaxIteration(10);
    ridgeLogisticOptimizer.optimize();
//        if (logger.isDebugEnabled()){
//            logger.debug("for cluster "+clusterIndex+" label "+labelIndex+" history= "+ridgeLogisticOptimizer.getOptimizer().getTerminator().getHistory());
//        }
}

Example 14 with LogisticRegression

use of edu.neu.ccs.pyramid.classification.logistic_regression.LogisticRegression in project pyramid by cheng-li.

the class ClassificationSynthesizerTest method test1.

private static void test1() {
    ClassificationSynthesizer classificationSynthesizer = ClassificationSynthesizer.getBuilder().setNumDataPoints(1000).setNumFeatures(2).setNoiseSD(0.00000001).build();
    ClfDataSet trainSet = classificationSynthesizer.multivarLine();
    ClfDataSet testSet = classificationSynthesizer.multivarLine();
    TRECFormat.save(trainSet, new File(TMP, "line1/train.trec"));
    TRECFormat.save(testSet, new File(TMP, "line1/test.trec"));
    RidgeLogisticTrainer trainer = RidgeLogisticTrainer.getBuilder().setGaussianPriorVariance(1).build();
    LogisticRegression logisticRegression = trainer.train(trainSet);
    System.out.println(Accuracy.accuracy(logisticRegression, trainSet));
    System.out.println(Accuracy.accuracy(logisticRegression, testSet));
    System.out.println(logisticRegression.getWeights().getWeightsForClass(0));
}

Example 15 with LogisticRegression

use of edu.neu.ccs.pyramid.classification.logistic_regression.LogisticRegression in project pyramid by cheng-li.

the class ClassificationSynthesizerTest method test2.

private static void test2() {
    ClassificationSynthesizer classificationSynthesizer = ClassificationSynthesizer.getBuilder().setNumDataPoints(100).setNumFeatures(2).setNoiseSD(0.00000001).build();
    ClfDataSet trainSet = classificationSynthesizer.multivarLine();
    ClfDataSet testSet = classificationSynthesizer.multivarLine();
    TRECFormat.save(trainSet, new File(TMP, "line2/train.trec"));
    TRECFormat.save(testSet, new File(TMP, "line2/test.trec"));
    RidgeLogisticTrainer trainer = RidgeLogisticTrainer.getBuilder().setGaussianPriorVariance(1).build();
    LogisticRegression logisticRegression = trainer.train(trainSet);
    System.out.println(Accuracy.accuracy(logisticRegression, trainSet));
    System.out.println(Accuracy.accuracy(logisticRegression, testSet));
    System.out.println(logisticRegression.getWeights().getWeightsForClass(0));
}