Example usage for org.apache.mahout.classifier.sgd OnlineLogisticRegression alpha

Introduction

In this page you can find the example usage for org.apache.mahout.classifier.sgd OnlineLogisticRegression alpha.

Prototype

public OnlineLogisticRegression alpha(double alpha) 

Source Link

Document

Chainable configuration option.

Usage

From source file:opennlp.addons.mahout.OnlineLogisticRegressionTrainer.java

License:Apache License

@Override
public MaxentModel doTrain(DataIndexer indexer) throws IOException {

    // TODO: Lets use the predMap here as well for encoding
    int numberOfOutcomes = indexer.getOutcomeLabels().length;
    int numberOfFeatures = indexer.getPredLabels().length;

    // TODO: Make these parameters configurable ...
    OnlineLogisticRegression pa = new OnlineLogisticRegression(numberOfOutcomes, numberOfFeatures, new L1());

    pa.alpha(1).stepOffset(250).decayExponent(0.9).lambda(3.0e-5).learningRate(3000);

    for (int k = 0; k < iterations; k++) {
        trainOnlineLearner(indexer, pa);

        // What should be reported at the end of every iteration ?!
        System.out.println("Iteration " + (k + 1));
    }//from  w  w w  .j ava2s.c  o m

    pa.close();

    return new VectorClassifierModel(pa, indexer.getOutcomeLabels(), createPrepMap(indexer));
}

---

From source file:org.wso2.siddhi.extension.ModelInitializer.java

License:Open Source License

public static OnlineLogisticRegression InitializeLogisticRegression(String modelPath) {
    OnlineLogisticRegression LRmodel = null;
    FileInputStream fileInputStream = null;
    ObjectInputStream objectInputStream = null;
    double[][] modelWeights = null;
    LogisticRegressionModel LRmodelObject;
    try {/* w w w  .ja v  a 2s  . com*/
        // get the values for hyper-parameters from model file.
        fileInputStream = new FileInputStream(modelPath);
        objectInputStream = new ObjectInputStream(fileInputStream);
        LRmodelObject = (LogisticRegressionModel) objectInputStream.readObject();
        LRmodel = new OnlineLogisticRegression(LRmodelObject.getNumCategories(), LRmodelObject.getNumFeatures(),
                new L2(1));
        LRmodel.learningRate(LRmodelObject.getLearningRate());
        LRmodel.lambda(LRmodelObject.getLambda());
        LRmodel.alpha(LRmodelObject.getAlpha());
        LRmodel.stepOffset(LRmodelObject.getStepOffset());
        LRmodel.decayExponent(LRmodelObject.getDecayExponent());
        modelWeights = LRmodelObject.getWeights();
        fileInputStream.close();
        objectInputStream.close();
        for (int i = 0; i < modelWeights.length; i++) {
            for (int j = 0; j < modelWeights[0].length; j++) {
                LRmodel.setBeta(i, j, modelWeights[i][j]);
            }
        }
    } catch (Exception e) {
        logger.error("Failed to create a Logistic Regression model from the file \"" + modelPath + "\"", e);
    } finally {
        try {
            fileInputStream.close();
            objectInputStream.close();
        } catch (IOException e) {
            logger.error("Failed to close the model input stream!", e);
        }
    }
    logger.info("Logistic Regression model execution plan successfully intialized for \"" + modelPath
            + "\" model file.");
    return LRmodel;
}

---