1 /* 2 * Licensed to the Apache Software Foundation (ASF) under one or more 3 * contributor license agreements. See the NOTICE file distributed with 4 * this work for additional information regarding copyright ownership. 5 * The ASF licenses this file to You under the Apache License, Version 2.0 6 * (the "License"); you may not use this file except in compliance with 7 * the License. You may obtain a copy of the License at 8 * 9 * http://www.apache.org/licenses/LICENSE-2.0 10 * 11 * Unless required by applicable law or agreed to in writing, software 12 * distributed under the License is distributed on an "AS IS" BASIS, 13 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 * See the License for the specific language governing permissions and 15 * limitations under the License. 16 */ 17 18 package org.apache.spark.examples.mllib; 19 20 // $example on$ 21 import java.util.Arrays; 22 23 import scala.Tuple2; 24 25 import org.apache.spark.api.java.*; 26 import org.apache.spark.api.java.function.Function; 27 import org.apache.spark.mllib.classification.LogisticRegressionModel; 28 import org.apache.spark.mllib.evaluation.BinaryClassificationMetrics; 29 import org.apache.spark.mllib.linalg.Vector; 30 import org.apache.spark.mllib.linalg.Vectors; 31 import org.apache.spark.mllib.optimization.*; 32 import org.apache.spark.mllib.regression.LabeledPoint; 33 import org.apache.spark.mllib.util.MLUtils; 34 import org.apache.spark.SparkConf; 35 import org.apache.spark.SparkContext; 36 // $example off$ 37 38 public class JavaLBFGSExample { main(String[] args)39 public static void main(String[] args) { 40 SparkConf conf = new SparkConf().setAppName("L-BFGS Example"); 41 SparkContext sc = new SparkContext(conf); 42 43 // $example on$ 44 String path = "data/mllib/sample_libsvm_data.txt"; 45 JavaRDD<LabeledPoint> data = MLUtils.loadLibSVMFile(sc, path).toJavaRDD(); 46 int numFeatures = data.take(1).get(0).features().size(); 47 48 // Split initial RDD into two... [60% training data, 40% testing data]. 49 JavaRDD<LabeledPoint> trainingInit = data.sample(false, 0.6, 11L); 50 JavaRDD<LabeledPoint> test = data.subtract(trainingInit); 51 52 // Append 1 into the training data as intercept. 53 JavaRDD<Tuple2<Object, Vector>> training = data.map( 54 new Function<LabeledPoint, Tuple2<Object, Vector>>() { 55 public Tuple2<Object, Vector> call(LabeledPoint p) { 56 return new Tuple2<Object, Vector>(p.label(), MLUtils.appendBias(p.features())); 57 } 58 }); 59 training.cache(); 60 61 // Run training algorithm to build the model. 62 int numCorrections = 10; 63 double convergenceTol = 1e-4; 64 int maxNumIterations = 20; 65 double regParam = 0.1; 66 Vector initialWeightsWithIntercept = Vectors.dense(new double[numFeatures + 1]); 67 68 Tuple2<Vector, double[]> result = LBFGS.runLBFGS( 69 training.rdd(), 70 new LogisticGradient(), 71 new SquaredL2Updater(), 72 numCorrections, 73 convergenceTol, 74 maxNumIterations, 75 regParam, 76 initialWeightsWithIntercept); 77 Vector weightsWithIntercept = result._1(); 78 double[] loss = result._2(); 79 80 final LogisticRegressionModel model = new LogisticRegressionModel( 81 Vectors.dense(Arrays.copyOf(weightsWithIntercept.toArray(), weightsWithIntercept.size() - 1)), 82 (weightsWithIntercept.toArray())[weightsWithIntercept.size() - 1]); 83 84 // Clear the default threshold. 85 model.clearThreshold(); 86 87 // Compute raw scores on the test set. 88 JavaRDD<Tuple2<Object, Object>> scoreAndLabels = test.map( 89 new Function<LabeledPoint, Tuple2<Object, Object>>() { 90 public Tuple2<Object, Object> call(LabeledPoint p) { 91 Double score = model.predict(p.features()); 92 return new Tuple2<Object, Object>(score, p.label()); 93 } 94 }); 95 96 // Get evaluation metrics. 97 BinaryClassificationMetrics metrics = 98 new BinaryClassificationMetrics(scoreAndLabels.rdd()); 99 double auROC = metrics.areaUnderROC(); 100 101 System.out.println("Loss of each step in training process"); 102 for (double l : loss) 103 System.out.println(l); 104 System.out.println("Area under ROC = " + auROC); 105 // $example off$ 106 } 107 } 108 109