layout: global title: “Migration Guide: MLlib (Machine Learning)” displayTitle: “Migration Guide: MLlib (Machine Learning)” license: | Licensed to the Apache Software Foundation (ASF) under one or more contributor license agreements. See the NOTICE file distributed with this work for additional information regarding copyright ownership. The ASF licenses this file to You under the Apache License, Version 2.0 (the “License”); you may not use this file except in compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Note that this migration guide describes the items specific to MLlib. Many items of SQL migration can be applied when migrating MLlib to higher versions for DataFrame-based APIs. Please refer Migration Guide: SQL, Datasets and DataFrame.
{:.no_toc}
OneHotEncoder which is deprecated in 2.3, is removed in 3.0 and OneHotEncoderEstimator is now renamed to OneHotEncoder.org.apache.spark.ml.image.ImageSchema.readImages which is deprecated in 2.3, is removed in 3.0, use spark.read.format('image') instead.org.apache.spark.mllib.clustering.KMeans.train with param Int runs which is deprecated in 2.1, is removed in 3.0. Use train method without runs instead.org.apache.spark.mllib.classification.LogisticRegressionWithSGD which is deprecated in 2.0, is removed in 3.0, use org.apache.spark.ml.classification.LogisticRegression or spark.mllib.classification.LogisticRegressionWithLBFGS instead.org.apache.spark.mllib.feature.ChiSqSelectorModel.isSorted which is deprecated in 2.1, is removed in 3.0, is not intended for subclasses to use.org.apache.spark.mllib.regression.RidgeRegressionWithSGD which is deprecated in 2.0, is removed in 3.0, use org.apache.spark.ml.regression.LinearRegression with elasticNetParam = 0.0. Note the default regParam is 0.01 for RidgeRegressionWithSGD, but is 0.0 for LinearRegression.org.apache.spark.mllib.regression.LassoWithSGD which is deprecated in 2.0, is removed in 3.0, use org.apache.spark.ml.regression.LinearRegression with elasticNetParam = 1.0. Note the default regParam is 0.01 for LassoWithSGD, but is 0.0 for LinearRegression.org.apache.spark.mllib.regression.LinearRegressionWithSGD which is deprecated in 2.0, is removed in 3.0, use org.apache.spark.ml.regression.LinearRegression or LBFGS instead.org.apache.spark.mllib.clustering.KMeans.getRuns and setRuns which are deprecated in 2.1, are removed in 3.0, have no effect since Spark 2.0.0.org.apache.spark.ml.LinearSVCModel.setWeightCol which is deprecated in 2.4, is removed in 3.0, is not intended for users.org.apache.spark.ml.classification.MultilayerPerceptronClassificationModel extends MultilayerPerceptronParams to expose the training params. As a result, layers in MultilayerPerceptronClassificationModel has been changed from Array[Int] to IntArrayParam. Users should use MultilayerPerceptronClassificationModel.getLayers instead of MultilayerPerceptronClassificationModel.layers to retrieve the size of layers.org.apache.spark.ml.classification.GBTClassifier.numTrees which is deprecated in 2.4.5, is removed in 3.0, use getNumTrees instead.org.apache.spark.ml.clustering.KMeansModel.computeCost which is deprecated in 2.4, is removed in 3.0, use ClusteringEvaluator instead.precision in org.apache.spark.mllib.evaluation.MulticlassMetrics which is deprecated in 2.0, is removed in 3.0. Use accuracy instead.recall in org.apache.spark.mllib.evaluation.MulticlassMetrics which is deprecated in 2.0, is removed in 3.0. Use accuracy instead.fMeasure in org.apache.spark.mllib.evaluation.MulticlassMetrics which is deprecated in 2.0, is removed in 3.0. Use accuracy instead.org.apache.spark.ml.util.GeneralMLWriter.context which is deprecated in 2.0, is removed in 3.0, use session instead.org.apache.spark.ml.util.MLWriter.context which is deprecated in 2.0, is removed in 3.0, use session instead.org.apache.spark.ml.util.MLReader.context which is deprecated in 2.0, is removed in 3.0, use session instead.abstract class UnaryTransformer[IN, OUT, T <: UnaryTransformer[IN, OUT, T]] is changed to abstract class UnaryTransformer[IN: TypeTag, OUT: TypeTag, T <: UnaryTransformer[IN, OUT, T]] in 3.0.{:.no_toc}
Deprecations
labels in StringIndexerModel is deprecated and will be removed in 3.1.0. Use labelsArray instead.computeCost in BisectingKMeansModel is deprecated and will be removed in future versions. Use ClusteringEvaluator instead.Changes of behavior
frequencyDesc or frequencyAsc as stringOrderType param in StringIndexer, in case of equal frequency, the order of strings is undefined. Since Spark 3.0, the strings with equal frequency are further sorted by alphabet. And since Spark 3.0, StringIndexer supports encoding multiple columns.Imputer requires input column to be Double or Float. In 3.0, this restriction is lifted so Imputer can handle all numeric types.HashingTF Transformer uses a corrected implementation of the murmur3 hash function to hash elements to vectors. HashingTF in Spark 3.0 will map elements to different positions in vectors than in Spark 2. However, HashingTF created with Spark 2.x and loaded with Spark 3.0 will still use the previous hash function and will not change behavior.setClassifier method in PySpark's OneVsRestModel has been removed in 3.0 for parity with the Scala implementation. Callers should not need to set the classifier in the model after creation.RandomForestRegressionModel doesn't update the parameter maps of the DecisionTreeRegressionModels underneath. This is fixed in 3.0.{:.no_toc}
LogisticRegressionTrainingSummary to a BinaryLogisticRegressionTrainingSummary. Users should instead use the model.binarySummary method. See SPARK-17139 for more detail (note this is an Experimental API). This does not affect the Python summary method, which will still work correctly for both multinomial and binary cases.{:.no_toc}
Deprecations
OneHotEncoder has been deprecated and will be removed in 3.0. It has been replaced by the new OneHotEncoderEstimator (see SPARK-13030). Note that OneHotEncoderEstimator will be renamed to OneHotEncoder in 3.0 (but OneHotEncoderEstimator will be kept as an alias).Changes of behavior
OneVsRest is now set to 1 (i.e. serial). In 2.2 and earlier versions, the level of parallelism was set to the default threadpool size in Scala.Word2Vec was incorrect when numIterations was set greater than 1. This will cause training results to be different between 2.3 and earlier versions.RFormula without an intercept were inconsistent with the output in R. This may change results from model training in this scenario.{:.no_toc}
There are no breaking changes.
{:.no_toc}
Deprecations
There are no deprecations.
Changes of behavior
regParam changed from 1.0 to 0.1 for ALS.train method (marked DeveloperApi). Note this does not affect the ALS Estimator or Model, nor MLlib's ALS class.Param.copy method.StringIndexer now handles NULL values in the same way as unseen values. Previously an exception would always be thrown regardless of the setting of the handleInvalid parameter.{:.no_toc}
Deprecated methods removed
setLabelCol in feature.ChiSqSelectorModelnumTrees in classification.RandomForestClassificationModel (This now refers to the Param called numTrees)numTrees in regression.RandomForestRegressionModel (This now refers to the Param called numTrees)model in regression.LinearRegressionSummaryvalidateParams in PipelineStagevalidateParams in Evaluator{:.no_toc}
Deprecations
DecisionTreeClassificationModel, GBTClassificationModel, RandomForestClassificationModel, DecisionTreeRegressionModel, GBTRegressionModel and RandomForestRegressionModelChanges of behavior
ChiSqSelector which will likely change its result. Now ChiSquareSelector use pValue rather than raw statistic to select a fixed number of top features.KMeans returns potentially fewer than k cluster centers in cases where k distinct centroids aren‘t available or aren’t selected.KMeans reduces the default number of steps from 5 to 2 for the k-means|| initialization mode.{:.no_toc}
There were several breaking changes in Spark 2.0, which are outlined below.
Linear algebra classes for DataFrame-based APIs
Spark's linear algebra dependencies were moved to a new project, mllib-local (see SPARK-13944). As part of this change, the linear algebra classes were copied to a new package, spark.ml.linalg. The DataFrame-based APIs in spark.ml now depend on the spark.ml.linalg classes, leading to a few breaking changes, predominantly in various model classes (see SPARK-14810 for a full list).
Note: the RDD-based APIs in spark.mllib continue to depend on the previous package spark.mllib.linalg.
Converting vectors and matrices
While most pipeline components support backward compatibility for loading, some existing DataFrames and pipelines in Spark versions prior to 2.0, that contain vector or matrix columns, may need to be migrated to the new spark.ml vector and matrix types. Utilities for converting DataFrame columns from spark.mllib.linalg to spark.ml.linalg types (and vice versa) can be found in spark.mllib.util.MLUtils.
There are also utility methods available for converting single instances of vectors and matrices. Use the asML method on a mllib.linalg.Vector / mllib.linalg.Matrix for converting to ml.linalg types, and mllib.linalg.Vectors.fromML / mllib.linalg.Matrices.fromML for converting to mllib.linalg types.
{% highlight python %} from pyspark.mllib.util import MLUtils
convertedVecDF = MLUtils.convertVectorColumnsToML(vecDF) convertedMatrixDF = MLUtils.convertMatrixColumnsToML(matrixDF)
mlVec = mllibVec.asML() mlMat = mllibMat.asML() {% endhighlight %}
Refer to the MLUtils Python docs for further detail.
{% highlight scala %} import org.apache.spark.mllib.util.MLUtils
// convert DataFrame columns val convertedVecDF = MLUtils.convertVectorColumnsToML(vecDF) val convertedMatrixDF = MLUtils.convertMatrixColumnsToML(matrixDF) // convert a single vector or matrix val mlVec: org.apache.spark.ml.linalg.Vector = mllibVec.asML val mlMat: org.apache.spark.ml.linalg.Matrix = mllibMat.asML {% endhighlight %}
Refer to the MLUtils Scala docs for further detail.
{% highlight java %} import org.apache.spark.mllib.util.MLUtils; import org.apache.spark.sql.Dataset;
// convert DataFrame columns Dataset convertedVecDF = MLUtils.convertVectorColumnsToML(vecDF); Dataset convertedMatrixDF = MLUtils.convertMatrixColumnsToML(matrixDF); // convert a single vector or matrix org.apache.spark.ml.linalg.Vector mlVec = mllibVec.asML(); org.apache.spark.ml.linalg.Matrix mlMat = mllibMat.asML(); {% endhighlight %}
Refer to the MLUtils Java docs for further detail.
Deprecated methods removed
Several deprecated methods were removed in the spark.mllib and spark.ml packages:
setScoreCol in ml.evaluation.BinaryClassificationEvaluatorweights in LinearRegression and LogisticRegression in spark.mlsetMaxNumIterations in mllib.optimization.LBFGS (marked as DeveloperApi)treeReduce and treeAggregate in mllib.rdd.RDDFunctions (these functions are available on RDDs directly, and were marked as DeveloperApi)defaultStrategy in mllib.tree.configuration.Strategybuild in mllib.tree.Nodemllib.util.MLUtilsA full list of breaking changes can be found at SPARK-14810.
{:.no_toc}
Deprecations
Deprecations in the spark.mllib and spark.ml packages include:
spark.ml.regression.LinearRegressionSummary, the model field has been deprecated.spark.ml.regression.RandomForestRegressionModel and spark.ml.classification.RandomForestClassificationModel, the numTrees parameter has been deprecated in favor of getNumTrees method.spark.ml.param.Params, the validateParams method has been deprecated. We move all functionality in overridden methods to the corresponding transformSchema.spark.mllib package, LinearRegressionWithSGD, LassoWithSGD, RidgeRegressionWithSGD and LogisticRegressionWithSGD have been deprecated. We encourage users to use spark.ml.regression.LinearRegression and spark.ml.classification.LogisticRegression.spark.mllib.evaluation.MulticlassMetrics, the parameters precision, recall and fMeasure have been deprecated in favor of accuracy.spark.ml.util.MLReader and spark.ml.util.MLWriter, the context method has been deprecated in favor of session.spark.ml.feature.ChiSqSelectorModel, the setLabelCol method has been deprecated since it was not used by ChiSqSelectorModel.Changes of behavior
Changes of behavior in the spark.mllib and spark.ml packages include:
spark.mllib.classification.LogisticRegressionWithLBFGS directly calls spark.ml.classification.LogisticRegression for binary classification now. This will introduce the following behavior changes for spark.mllib.classification.LogisticRegressionWithLBFGS:spark.ml.classification.LogisticRegression, the default value of spark.mllib.classification.LogisticRegressionWithLBFGS: convergenceTol has been changed from 1E-4 to 1E-6.PowerIterationClustering which will likely change its result.LDA using the EM optimizer will keep the last checkpoint by default, if checkpointing is being used.Word2Vec now respects sentence boundaries. Previously, it did not handle them correctly.HashingTF uses MurmurHash3 as default hash algorithm in both spark.ml and spark.mllib.expectedType argument for PySpark Param was removed.Param values, which were mismatched between pipelines in Scala and Python, have been changed.QuantileDiscretizer now uses spark.sql.DataFrameStatFunctions.approxQuantile to find splits (previously used custom sampling logic). The output buckets will differ for same input data and params.There are no breaking API changes in the spark.mllib or spark.ml packages, but there are deprecations and changes of behavior.
Deprecations:
spark.mllib.clustering.KMeans, the runs parameter has been deprecated.spark.ml.classification.LogisticRegressionModel and spark.ml.regression.LinearRegressionModel, the weights field has been deprecated in favor of the new name coefficients. This helps disambiguate from instance (row) “weights” given to algorithms.Changes of behavior:
spark.mllib.tree.GradientBoostedTrees: validationTol has changed semantics in 1.6. Previously, it was a threshold for absolute change in error. Now, it resembles the behavior of GradientDescent's convergenceTol: For large errors, it uses relative error (relative to the previous error); for small errors (< 0.01), it uses absolute error.spark.ml.feature.RegexTokenizer: Previously, it did not convert strings to lowercase before tokenizing. Now, it converts to lowercase by default, with an option not to. This matches the behavior of the simpler Tokenizer transformer.In the spark.mllib package, there are no breaking API changes but several behavior changes:
RegressionMetrics.explainedVariance returns the average regression sum of squares.NaiveBayesModel.labels become sorted.GradientDescent has a default convergence tolerance 1e-3, and hence iterations might end earlier than 1.4.In the spark.ml package, there exists one breaking API change and one behavior change:
Params.setDefault due to a Scala compiler bug.Evaluator.isLargerBetter is added to indicate metric ordering. Metrics like RMSE no longer flip signs as in 1.4.In the spark.mllib package, there were several breaking changes, but all in DeveloperApi or Experimental APIs:
Loss.gradient method was changed. This is only an issues for users who wrote their own losses for GBTs.apply and copy methods for the case class BoostingStrategy have been changed because of a modification to the case class fields. This could be an issue for users who use BoostingStrategy to set GBT parameters.LDA.run has changed. It now returns an abstract class LDAModel instead of the concrete class DistributedLDAModel. The object of type LDAModel can still be cast to the appropriate concrete type, which depends on the optimization algorithm.In the spark.ml package, several major API changes occurred, including:
Param and other APIs for specifying parametersuid unique IDs for Pipeline componentsSince the spark.ml API was an alpha component in Spark 1.3, we do not list all changes here. However, since 1.4 spark.ml is no longer an alpha component, we will provide details on any API changes for future releases.
In the spark.mllib package, there were several breaking changes. The first change (in ALS) is the only one in a component not marked as Alpha or Experimental.
ALS, the extraneous method solveLeastSquares has been removed. The DeveloperApi method analyzeBlocks was also removed.StandardScalerModel remains an Alpha component. In it, the variance method has been replaced with the std method. To compute the column variance values returned by the original variance method, simply square the standard deviation values returned by std.StreamingLinearRegressionWithSGD remains an Experimental component. In it, there were two changes:model is no longer public.DecisionTree remains an Experimental component. In it and its associated classes, there were several changes:DecisionTree, the deprecated class method train has been removed. (The object/static train methods remain.)Strategy, the checkpointDir parameter has been removed. Checkpointing is still supported, but the checkpoint directory must be set before calling tree and tree ensemble training.PythonMLlibAPI (the interface between Scala/Java and Python for MLlib) was a public API but is now private, declared private[python]. This was never meant for external use.In the spark.ml package, the main API changes are from Spark SQL. We list the most important changes here:
spark.ml which used to use SchemaRDD now use DataFrame.RDDs of LabeledPoint into SchemaRDDs by calling import sqlContext._ where sqlContext was an instance of SQLContext. These implicits have been moved, so we now call import sqlContext.implicits._.Other changes were in LogisticRegression:
scoreCol output column (with default value “score”) was renamed to be probabilityCol (with default value “probability”). The type was originally Double (for the probability of class 1.0), but it is now Vector (for the probability of each class, to support multiclass classification in the future).LogisticRegressionModel did not include an intercept. In Spark 1.3, it includes an intercept; however, it will always be 0.0 since it uses the default settings for spark.mllib.LogisticRegressionWithLBFGS. The option to use an intercept will be added in the future.The only API changes in MLlib v1.2 are in DecisionTree, which continues to be an experimental API in MLlib 1.2:
(Breaking change) The Scala API for classification takes a named argument specifying the number of classes. In MLlib v1.1, this argument was called numClasses in Python and numClassesForClassification in Scala. In MLlib v1.2, the names are both set to numClasses. This numClasses parameter is specified either via Strategy or via DecisionTree static trainClassifier and trainRegressor methods.
(Breaking change) The API for Node has changed. This should generally not affect user code, unless the user manually constructs decision trees (instead of using the trainClassifier or trainRegressor methods). The tree Node now includes more information, including the probability of the predicted label (for classification).
Printing methods' output has changed. The toString (Scala/Java) and __repr__ (Python) methods used to print the full model; they now print a summary. For the full model, use toDebugString.
Examples in the Spark distribution and examples in the Decision Trees Guide have been updated accordingly.
The only API changes in MLlib v1.1 are in DecisionTree, which continues to be an experimental API in MLlib 1.1:
(Breaking change) The meaning of tree depth has been changed by 1 in order to match the implementations of trees in scikit-learn and in rpart. In MLlib v1.0, a depth-1 tree had 1 leaf node, and a depth-2 tree had 1 root node and 2 leaf nodes. In MLlib v1.1, a depth-0 tree has 1 leaf node, and a depth-1 tree has 1 root node and 2 leaf nodes. This depth is specified by the maxDepth parameter in Strategy or via DecisionTree static trainClassifier and trainRegressor methods.
(Non-breaking change) We recommend using the newly added trainClassifier and trainRegressor methods to build a DecisionTree, rather than using the old parameter class Strategy. These new training methods explicitly separate classification and regression, and they replace specialized parameter types with simple String types.
Examples of the new recommended trainClassifier and trainRegressor are given in the Decision Trees Guide.
In MLlib v1.0, we support both dense and sparse input in a unified way, which introduces a few breaking changes. If your data is sparse, please store it in a sparse format instead of dense to take advantage of sparsity in both storage and computation. Details are described below.