From e0833c5958bbd73ff27cfe6865648d7b6e5a99bc Mon Sep 17 00:00:00 2001 From: Xiangrui Meng Date: Mon, 4 May 2015 11:28:59 -0700 Subject: [SPARK-5956] [MLLIB] Pipeline components should be copyable. This PR added `copy(extra: ParamMap): Params` to `Params`, which makes a copy of the current instance with a randomly generated uid and some extra param values. With this change, we only need to implement `fit` and `transform` without extra param values given the default implementation of `fit(dataset, extra)`: ~~~scala def fit(dataset: DataFrame, extra: ParamMap): Model = { copy(extra).fit(dataset) } ~~~ Inside `fit` and `transform`, since only the embedded values are used, I added `$` as an alias for `getOrDefault` to make the code easier to read. For example, in `LinearRegression.fit` we have: ~~~scala val effectiveRegParam = $(regParam) / yStd val effectiveL1RegParam = $(elasticNetParam) * effectiveRegParam val effectiveL2RegParam = (1.0 - $(elasticNetParam)) * effectiveRegParam ~~~ Meta-algorithm like `Pipeline` implements its own `copy(extra)`. So the fitted pipeline model stored all copied stages (no matter whether it is a transformer or a model). Other changes: * `Params$.inheritValues` is moved to `Params!.copyValues` and returns the target instance. * `fittingParamMap` was removed because the `parent` carries this information. * `validate` was renamed to `validateParams` to be more precise. TODOs: * [x] add tests for newly added methods * [ ] update documentation jkbradley dbtsai Author: Xiangrui Meng Closes #5820 from mengxr/SPARK-5956 and squashes the following commits: 7bef88d [Xiangrui Meng] address comments 05229c3 [Xiangrui Meng] assert -> assertEquals b2927b1 [Xiangrui Meng] organize imports f14456b [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-5956 93e7924 [Xiangrui Meng] add tests for hasParam & copy 463ecae [Xiangrui Meng] merge master 2b954c3 [Xiangrui Meng] update Binarizer 465dd12 [Xiangrui Meng] Merge remote-tracking branch 'apache/master' into SPARK-5956 282a1a8 [Xiangrui Meng] fix test 819dd2d [Xiangrui Meng] merge master b642872 [Xiangrui Meng] example code runs 5a67779 [Xiangrui Meng] examples compile c76b4d1 [Xiangrui Meng] fix all unit tests 0f4fd64 [Xiangrui Meng] fix some tests 9286a22 [Xiangrui Meng] copyValues to trained models 53e0973 [Xiangrui Meng] move inheritValues to Params and rename it to copyValues 9ee004e [Xiangrui Meng] merge copy and copyWith; rename validate to validateParams d882afc [Xiangrui Meng] test compile f082a31 [Xiangrui Meng] make Params copyable and simply handling of extra params in all spark.ml components --- .../spark/examples/ml/JavaDeveloperApiExample.java | 24 +++++++--------------- .../spark/examples/ml/JavaSimpleParamsExample.java | 4 ++-- .../spark/examples/ml/DecisionTreeExample.scala | 6 ++---- .../spark/examples/ml/DeveloperApiExample.scala | 22 ++++++++------------ .../org/apache/spark/examples/ml/GBTExample.scala | 4 ++-- .../spark/examples/ml/RandomForestExample.scala | 6 ++---- .../spark/examples/ml/SimpleParamsExample.scala | 4 ++-- 7 files changed, 25 insertions(+), 45 deletions(-) (limited to 'examples') diff --git a/examples/src/main/java/org/apache/spark/examples/ml/JavaDeveloperApiExample.java b/examples/src/main/java/org/apache/spark/examples/ml/JavaDeveloperApiExample.java index 46377a99c4..eac4f898a4 100644 --- a/examples/src/main/java/org/apache/spark/examples/ml/JavaDeveloperApiExample.java +++ b/examples/src/main/java/org/apache/spark/examples/ml/JavaDeveloperApiExample.java @@ -28,7 +28,6 @@ import org.apache.spark.ml.classification.Classifier; import org.apache.spark.ml.classification.ClassificationModel; import org.apache.spark.ml.param.IntParam; import org.apache.spark.ml.param.ParamMap; -import org.apache.spark.ml.param.Params$; import org.apache.spark.mllib.linalg.BLAS; import org.apache.spark.mllib.linalg.Vector; import org.apache.spark.mllib.linalg.Vectors; @@ -129,16 +128,16 @@ class MyJavaLogisticRegression // This method is used by fit(). // In Java, we have to make it public since Java does not understand Scala's protected modifier. - public MyJavaLogisticRegressionModel train(DataFrame dataset, ParamMap paramMap) { + public MyJavaLogisticRegressionModel train(DataFrame dataset) { // Extract columns from data using helper method. - JavaRDD oldDataset = extractLabeledPoints(dataset, paramMap).toJavaRDD(); + JavaRDD oldDataset = extractLabeledPoints(dataset).toJavaRDD(); // Do learning to estimate the weight vector. int numFeatures = oldDataset.take(1).get(0).features().size(); Vector weights = Vectors.zeros(numFeatures); // Learning would happen here. // Create a model, and return it. - return new MyJavaLogisticRegressionModel(this, paramMap, weights); + return new MyJavaLogisticRegressionModel(this, weights); } } @@ -155,18 +154,11 @@ class MyJavaLogisticRegressionModel private MyJavaLogisticRegression parent_; public MyJavaLogisticRegression parent() { return parent_; } - private ParamMap fittingParamMap_; - public ParamMap fittingParamMap() { return fittingParamMap_; } - private Vector weights_; public Vector weights() { return weights_; } - public MyJavaLogisticRegressionModel( - MyJavaLogisticRegression parent_, - ParamMap fittingParamMap_, - Vector weights_) { + public MyJavaLogisticRegressionModel(MyJavaLogisticRegression parent_, Vector weights_) { this.parent_ = parent_; - this.fittingParamMap_ = fittingParamMap_; this.weights_ = weights_; } @@ -210,10 +202,8 @@ class MyJavaLogisticRegressionModel * In Java, we have to make this method public since Java does not understand Scala's protected * modifier. */ - public MyJavaLogisticRegressionModel copy() { - MyJavaLogisticRegressionModel m = - new MyJavaLogisticRegressionModel(parent_, fittingParamMap_, weights_); - Params$.MODULE$.inheritValues(this.extractParamMap(), this, m); - return m; + @Override + public MyJavaLogisticRegressionModel copy(ParamMap extra) { + return copyValues(new MyJavaLogisticRegressionModel(parent_, weights_), extra); } } diff --git a/examples/src/main/java/org/apache/spark/examples/ml/JavaSimpleParamsExample.java b/examples/src/main/java/org/apache/spark/examples/ml/JavaSimpleParamsExample.java index 4e02acce69..29158d5c85 100644 --- a/examples/src/main/java/org/apache/spark/examples/ml/JavaSimpleParamsExample.java +++ b/examples/src/main/java/org/apache/spark/examples/ml/JavaSimpleParamsExample.java @@ -71,7 +71,7 @@ public class JavaSimpleParamsExample { // we can view the parameters it used during fit(). // This prints the parameter (name: value) pairs, where names are unique IDs for this // LogisticRegression instance. - System.out.println("Model 1 was fit using parameters: " + model1.fittingParamMap()); + System.out.println("Model 1 was fit using parameters: " + model1.parent().extractParamMap()); // We may alternatively specify parameters using a ParamMap. ParamMap paramMap = new ParamMap(); @@ -87,7 +87,7 @@ public class JavaSimpleParamsExample { // Now learn a new model using the paramMapCombined parameters. // paramMapCombined overrides all parameters set earlier via lr.set* methods. LogisticRegressionModel model2 = lr.fit(training, paramMapCombined); - System.out.println("Model 2 was fit using parameters: " + model2.fittingParamMap()); + System.out.println("Model 2 was fit using parameters: " + model2.parent().extractParamMap()); // Prepare test documents. List localTest = Lists.newArrayList( diff --git a/examples/src/main/scala/org/apache/spark/examples/ml/DecisionTreeExample.scala b/examples/src/main/scala/org/apache/spark/examples/ml/DecisionTreeExample.scala index 9002e99d82..8340d91101 100644 --- a/examples/src/main/scala/org/apache/spark/examples/ml/DecisionTreeExample.scala +++ b/examples/src/main/scala/org/apache/spark/examples/ml/DecisionTreeExample.scala @@ -276,16 +276,14 @@ object DecisionTreeExample { // Get the trained Decision Tree from the fitted PipelineModel algo match { case "classification" => - val treeModel = pipelineModel.getModel[DecisionTreeClassificationModel]( - dt.asInstanceOf[DecisionTreeClassifier]) + val treeModel = pipelineModel.stages.last.asInstanceOf[DecisionTreeClassificationModel] if (treeModel.numNodes < 20) { println(treeModel.toDebugString) // Print full model. } else { println(treeModel) // Print model summary. } case "regression" => - val treeModel = pipelineModel.getModel[DecisionTreeRegressionModel]( - dt.asInstanceOf[DecisionTreeRegressor]) + val treeModel = pipelineModel.stages.last.asInstanceOf[DecisionTreeRegressionModel] if (treeModel.numNodes < 20) { println(treeModel.toDebugString) // Print full model. } else { diff --git a/examples/src/main/scala/org/apache/spark/examples/ml/DeveloperApiExample.scala b/examples/src/main/scala/org/apache/spark/examples/ml/DeveloperApiExample.scala index 2245fa429f..2a2d067727 100644 --- a/examples/src/main/scala/org/apache/spark/examples/ml/DeveloperApiExample.scala +++ b/examples/src/main/scala/org/apache/spark/examples/ml/DeveloperApiExample.scala @@ -18,13 +18,12 @@ package org.apache.spark.examples.ml import org.apache.spark.{SparkConf, SparkContext} -import org.apache.spark.ml.classification.{Classifier, ClassifierParams, ClassificationModel} -import org.apache.spark.ml.param.{Params, IntParam, ParamMap} +import org.apache.spark.ml.classification.{ClassificationModel, Classifier, ClassifierParams} +import org.apache.spark.ml.param.{IntParam, ParamMap} import org.apache.spark.mllib.linalg.{BLAS, Vector, Vectors} import org.apache.spark.mllib.regression.LabeledPoint import org.apache.spark.sql.{DataFrame, Row, SQLContext} - /** * A simple example demonstrating how to write your own learning algorithm using Estimator, * Transformer, and other abstractions. @@ -99,7 +98,7 @@ private trait MyLogisticRegressionParams extends ClassifierParams { * class since the maxIter parameter is only used during training (not in the Model). */ val maxIter: IntParam = new IntParam(this, "maxIter", "max number of iterations") - def getMaxIter: Int = getOrDefault(maxIter) + def getMaxIter: Int = $(maxIter) } /** @@ -117,18 +116,16 @@ private class MyLogisticRegression def setMaxIter(value: Int): this.type = set(maxIter, value) // This method is used by fit() - override protected def train( - dataset: DataFrame, - paramMap: ParamMap): MyLogisticRegressionModel = { + override protected def train(dataset: DataFrame): MyLogisticRegressionModel = { // Extract columns from data using helper method. - val oldDataset = extractLabeledPoints(dataset, paramMap) + val oldDataset = extractLabeledPoints(dataset) // Do learning to estimate the weight vector. val numFeatures = oldDataset.take(1)(0).features.size val weights = Vectors.zeros(numFeatures) // Learning would happen here. // Create a model, and return it. - new MyLogisticRegressionModel(this, paramMap, weights) + new MyLogisticRegressionModel(this, weights) } } @@ -139,7 +136,6 @@ private class MyLogisticRegression */ private class MyLogisticRegressionModel( override val parent: MyLogisticRegression, - override val fittingParamMap: ParamMap, val weights: Vector) extends ClassificationModel[Vector, MyLogisticRegressionModel] with MyLogisticRegressionParams { @@ -176,9 +172,7 @@ private class MyLogisticRegressionModel( * * This is used for the default implementation of [[transform()]]. */ - override protected def copy(): MyLogisticRegressionModel = { - val m = new MyLogisticRegressionModel(parent, fittingParamMap, weights) - Params.inheritValues(extractParamMap(), this, m) - m + override def copy(extra: ParamMap): MyLogisticRegressionModel = { + copyValues(new MyLogisticRegressionModel(parent, weights), extra) } } diff --git a/examples/src/main/scala/org/apache/spark/examples/ml/GBTExample.scala b/examples/src/main/scala/org/apache/spark/examples/ml/GBTExample.scala index 5fccb142d4..c5899b6683 100644 --- a/examples/src/main/scala/org/apache/spark/examples/ml/GBTExample.scala +++ b/examples/src/main/scala/org/apache/spark/examples/ml/GBTExample.scala @@ -201,14 +201,14 @@ object GBTExample { // Get the trained GBT from the fitted PipelineModel algo match { case "classification" => - val rfModel = pipelineModel.getModel[GBTClassificationModel](dt.asInstanceOf[GBTClassifier]) + val rfModel = pipelineModel.stages.last.asInstanceOf[GBTClassificationModel] if (rfModel.totalNumNodes < 30) { println(rfModel.toDebugString) // Print full model. } else { println(rfModel) // Print model summary. } case "regression" => - val rfModel = pipelineModel.getModel[GBTRegressionModel](dt.asInstanceOf[GBTRegressor]) + val rfModel = pipelineModel.stages.last.asInstanceOf[GBTRegressionModel] if (rfModel.totalNumNodes < 30) { println(rfModel.toDebugString) // Print full model. } else { diff --git a/examples/src/main/scala/org/apache/spark/examples/ml/RandomForestExample.scala b/examples/src/main/scala/org/apache/spark/examples/ml/RandomForestExample.scala index 9b909324ec..7f88d2681b 100644 --- a/examples/src/main/scala/org/apache/spark/examples/ml/RandomForestExample.scala +++ b/examples/src/main/scala/org/apache/spark/examples/ml/RandomForestExample.scala @@ -209,16 +209,14 @@ object RandomForestExample { // Get the trained Random Forest from the fitted PipelineModel algo match { case "classification" => - val rfModel = pipelineModel.getModel[RandomForestClassificationModel]( - dt.asInstanceOf[RandomForestClassifier]) + val rfModel = pipelineModel.stages.last.asInstanceOf[RandomForestClassificationModel] if (rfModel.totalNumNodes < 30) { println(rfModel.toDebugString) // Print full model. } else { println(rfModel) // Print model summary. } case "regression" => - val rfModel = pipelineModel.getModel[RandomForestRegressionModel]( - dt.asInstanceOf[RandomForestRegressor]) + val rfModel = pipelineModel.stages.last.asInstanceOf[RandomForestRegressionModel] if (rfModel.totalNumNodes < 30) { println(rfModel.toDebugString) // Print full model. } else { diff --git a/examples/src/main/scala/org/apache/spark/examples/ml/SimpleParamsExample.scala b/examples/src/main/scala/org/apache/spark/examples/ml/SimpleParamsExample.scala index bf805149d0..e8a991f50e 100644 --- a/examples/src/main/scala/org/apache/spark/examples/ml/SimpleParamsExample.scala +++ b/examples/src/main/scala/org/apache/spark/examples/ml/SimpleParamsExample.scala @@ -63,7 +63,7 @@ object SimpleParamsExample { // we can view the parameters it used during fit(). // This prints the parameter (name: value) pairs, where names are unique IDs for this // LogisticRegression instance. - println("Model 1 was fit using parameters: " + model1.fittingParamMap) + println("Model 1 was fit using parameters: " + model1.parent.extractParamMap()) // We may alternatively specify parameters using a ParamMap, // which supports several methods for specifying parameters. @@ -78,7 +78,7 @@ object SimpleParamsExample { // Now learn a new model using the paramMapCombined parameters. // paramMapCombined overrides all parameters set earlier via lr.set* methods. val model2 = lr.fit(training.toDF(), paramMapCombined) - println("Model 2 was fit using parameters: " + model2.fittingParamMap) + println("Model 2 was fit using parameters: " + model2.parent.extractParamMap()) // Prepare test data. val test = sc.parallelize(Seq( -- cgit v1.2.3