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.ChiSqSelectorModel
numTrees
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.LinearRegressionSummary
validateParams
in PipelineStage
validateParams
in Evaluator
{:.no_toc}
Deprecations
DecisionTreeClassificationModel
, GBTClassificationModel
, RandomForestClassificationModel
, DecisionTreeRegressionModel
, GBTRegressionModel
and RandomForestRegressionModel
Changes 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.BinaryClassificationEvaluator
weights
in LinearRegression
and LogisticRegression
in spark.ml
setMaxNumIterations
in mllib.optimization.LBFGS
(marked as DeveloperApi
)treeReduce
and treeAggregate
in mllib.rdd.RDDFunctions
(these functions are available on RDD
s directly, and were marked as DeveloperApi
)defaultStrategy
in mllib.tree.configuration.Strategy
build
in mllib.tree.Node
mllib.util.MLUtils
A 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.RDD
s of LabeledPoint
into SchemaRDD
s 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.