From 2be4255a05e7a1548f51b02f6bf62507f1c3414b Mon Sep 17 00:00:00 2001 From: Yanbo Liang Date: Wed, 25 Mar 2015 17:05:56 +0000 Subject: [PATCH] [SPARK-6496] [MLLIB] GeneralizedLinearAlgorithm.run(input, initialWeights) should initialize numFeatures In GeneralizedLinearAlgorithm ```numFeatures``` is default to -1, we need to update it to correct value when we call run() to train a model. ```LogisticRegressionWithLBFGS.run(input)``` works well, but when we call ```LogisticRegressionWithLBFGS.run(input, initialWeights)``` to train multiclass classification model, it will throw exception due to the numFeatures is not updated. In this PR, we just update numFeatures at the beginning of GeneralizedLinearAlgorithm.run(input, initialWeights) and add test case. Author: Yanbo Liang Closes #5167 from yanboliang/spark-6496 and squashes the following commits: 8131c48 [Yanbo Liang] LogisticRegressionWithLBFGS.run(input, initialWeights) should initialize numFeatures (cherry picked from commit 10c78607b2724f5a64b0cdb966e9c5805f23919b) Signed-off-by: Sean Owen --- .../spark/mllib/regression/GeneralizedLinearAlgorithm.scala | 4 ++++ .../mllib/classification/LogisticRegressionSuite.scala | 6 ++++++ 2 files changed, 10 insertions(+) diff --git a/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala b/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala index 7c66e8cdebdbe..9a2751aef908e 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala @@ -196,6 +196,10 @@ abstract class GeneralizedLinearAlgorithm[M <: GeneralizedLinearModel] */ def run(input: RDD[LabeledPoint], initialWeights: Vector): M = { + if (numFeatures < 0) { + numFeatures = input.map(_.features.size).first() + } + if (input.getStorageLevel == StorageLevel.NONE) { logWarning("The input data is not directly cached, which may hurt performance if its" + " parent RDDs are also uncached.") diff --git a/mllib/src/test/scala/org/apache/spark/mllib/classification/LogisticRegressionSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/classification/LogisticRegressionSuite.scala index aaa81da9e273c..a26c52852c4d7 100644 --- a/mllib/src/test/scala/org/apache/spark/mllib/classification/LogisticRegressionSuite.scala +++ b/mllib/src/test/scala/org/apache/spark/mllib/classification/LogisticRegressionSuite.scala @@ -425,6 +425,12 @@ class LogisticRegressionSuite extends FunSuite with MLlibTestSparkContext with M val model = lr.run(testRDD) + val numFeatures = testRDD.map(_.features.size).first() + val initialWeights = Vectors.dense(new Array[Double]((numFeatures + 1) * 2)) + val model2 = lr.run(testRDD, initialWeights) + + LogisticRegressionSuite.checkModelsEqual(model, model2) + /** * The following is the instruction to reproduce the model using R's glmnet package. *