[SPARK-23042][ML] Use OneHotEncoderModel to encode labels in MultilayerPerceptronClassifier

## What changes were proposed in this pull request?

In MultilayerPerceptronClassifier, we use RDD operation to encode labels for now. I think we should use ML's OneHotEncoderEstimator/Model to do the encoding.

## How was this patch tested?

Existing tests.

Closes #20232 from viirya/SPARK-23042.

Authored-by: Liang-Chi Hsieh <viirya@gmail.com>
Signed-off-by: DB Tsai <d_tsai@apple.com>

R/pkg/tests/fulltests/test_mllib_classification.R

@@ -382,10 +382,10 @@ test_that("spark.mlp", {
   trainidxs <- base::sample(nrow(data), nrow(data) * 0.7)
   traindf <- as.DataFrame(data[trainidxs, ])
   testdf <- as.DataFrame(rbind(data[-trainidxs, ], c(0, "the other")))
-  model <- spark.mlp(traindf, clicked ~ ., layers = c(1, 3))
+  model <- spark.mlp(traindf, clicked ~ ., layers = c(1, 2))
   predictions <- predict(model, testdf)
   expect_error(collect(predictions))
-  model <- spark.mlp(traindf, clicked ~ ., layers = c(1, 3), handleInvalid = "skip")
+  model <- spark.mlp(traindf, clicked ~ ., layers = c(1, 2), handleInvalid = "skip")
   predictions <- predict(model, testdf)
   expect_equal(class(collect(predictions)$clicked[1]), "list")
 

R/pkg/vignettes/sparkr-vignettes.Rmd

@@ -654,7 +654,7 @@ We use Titanic data set to show how to use `spark.mlp` in classification.
 t <- as.data.frame(Titanic)
 training <- createDataFrame(t)
 # fit a Multilayer Perceptron Classification Model
-model <- spark.mlp(training, Survived ~ Age + Sex, blockSize = 128, layers = c(2, 3), solver = "l-bfgs", maxIter = 100, tol = 0.5, stepSize = 1, seed = 1, initialWeights = c( 0, 0, 0, 5, 5, 5, 9, 9, 9))
+model <- spark.mlp(training, Survived ~ Age + Sex, blockSize = 128, layers = c(2, 2), solver = "l-bfgs", maxIter = 100, tol = 0.5, stepSize = 1, seed = 1, initialWeights = c( 0, 0, 5, 5, 9, 9))
 ```
 
 To avoid lengthy display, we only present partial results of the model summary. You can check the full result from your sparkR shell.

docs/sparkr.md

@@ -667,3 +667,7 @@ You can inspect the search path in R with [`search()`](https://stat.ethz.ch/R-ma
 ## Upgrading to SparkR 2.3.1 and above
 
  - In SparkR 2.3.0 and earlier, the `start` parameter of `substr` method was wrongly subtracted by one and considered as 0-based. This can lead to inconsistent substring results and also does not match with the behaviour with `substr` in R. In version 2.3.1 and later, it has been fixed so the `start` parameter of `substr` method is now 1-base. As an example, `substr(lit('abcdef'), 2, 4))` would result to `abc` in SparkR 2.3.0, and the result would be `bcd` in SparkR 2.3.1.
+
+## Upgrading to SparkR 2.4.0
+
+ - Previously, we don't check the validity of the size of the last layer in `spark.mlp`. For example, if the training data only has two labels, a `layers` param like `c(1, 3)` doesn't cause an error previously, now it does.

mllib/src/main/scala/org/apache/spark/ml/classification/MultilayerPerceptronClassifier.scala

@@ -23,13 +23,13 @@ import org.apache.hadoop.fs.Path
 
 import org.apache.spark.annotation.Since
 import org.apache.spark.ml.ann.{FeedForwardTopology, FeedForwardTrainer}
-import org.apache.spark.ml.feature.LabeledPoint
+import org.apache.spark.ml.feature.OneHotEncoderModel
 import org.apache.spark.ml.linalg.{Vector, Vectors}
 import org.apache.spark.ml.param._
 import org.apache.spark.ml.param.shared._
 import org.apache.spark.ml.util._
 import org.apache.spark.ml.util.Instrumentation.instrumented
-import org.apache.spark.sql.Dataset
+import org.apache.spark.sql.{Dataset, Row}
 
 /** Params for Multilayer Perceptron. */
 private[classification] trait MultilayerPerceptronParams extends ProbabilisticClassifierParams
@@ -103,36 +103,6 @@ private[classification] trait MultilayerPerceptronParams extends ProbabilisticCl
     solver -> LBFGS, stepSize -> 0.03)
 }
 
-/** Label to vector converter. */
-private object LabelConverter {
-  // TODO: Use OneHotEncoder instead
-  /**
-   * Encodes a label as a vector.
-   * Returns a vector of given length with zeroes at all positions
-   * and value 1.0 at the position that corresponds to the label.
-   *
-   * @param labeledPoint labeled point
-   * @param labelCount total number of labels
-   * @return pair of features and vector encoding of a label
-   */
-  def encodeLabeledPoint(labeledPoint: LabeledPoint, labelCount: Int): (Vector, Vector) = {
-    val output = Array.fill(labelCount)(0.0)
-    output(labeledPoint.label.toInt) = 1.0
-    (labeledPoint.features, Vectors.dense(output))
-  }
-
-  /**
-   * Converts a vector to a label.
-   * Returns the position of the maximal element of a vector.
-   *
-   * @param output label encoded with a vector
-   * @return label
-   */
-  def decodeLabel(output: Vector): Double = {
-    output.argmax.toDouble
-  }
-}
-
 /**
  * Classifier trainer based on the Multilayer Perceptron.
  * Each layer has sigmoid activation function, output layer has softmax.
@@ -243,8 +213,18 @@ class MultilayerPerceptronClassifier @Since("1.5.0") (
     instr.logNumClasses(labels)
     instr.logNumFeatures(myLayers.head)
 
-    val lpData = extractLabeledPoints(dataset)
+    // One-hot encoding for labels using OneHotEncoderModel.
-    val data = lpData.map(lp => LabelConverter.encodeLabeledPoint(lp, labels))
+    // As we already know the length of encoding, we skip fitting and directly create
+    // the model.
+    val encodedLabelCol = "_encoded" + $(labelCol)
+    val encodeModel = new OneHotEncoderModel(uid, Array(labels))
+      .setInputCols(Array($(labelCol)))
+      .setOutputCols(Array(encodedLabelCol))
+      .setDropLast(false)
+    val encodedDataset = encodeModel.transform(dataset)
+    val data = encodedDataset.select($(featuresCol), encodedLabelCol).rdd.map {
+      case Row(features: Vector, encodedLabel: Vector) => (features, encodedLabel)
+    }
     val topology = FeedForwardTopology.multiLayerPerceptron(myLayers, softmaxOnTop = true)
     val trainer = new FeedForwardTrainer(topology, myLayers(0), myLayers.last)
     if (isDefined(initialWeights)) {
@@ -323,7 +303,7 @@ class MultilayerPerceptronClassificationModel private[ml] (
    * This internal method is used to implement `transform()` and output [[predictionCol]].
    */
   override def predict(features: Vector): Double = {
-    LabelConverter.decodeLabel(mlpModel.predict(features))
+    mlpModel.predict(features).argmax.toDouble
   }
 
   @Since("1.5.0")

project/MimaExcludes.scala

@@ -97,7 +97,10 @@ object MimaExcludes {
     ProblemFilters.exclude[InheritedNewAbstractMethodProblem]("org.apache.spark.ml.param.shared.HasValidationIndicatorCol.validationIndicatorCol"),
     ProblemFilters.exclude[InheritedNewAbstractMethodProblem]("org.apache.spark.ml.param.shared.HasValidationIndicatorCol.getValidationIndicatorCol"),
     ProblemFilters.exclude[InheritedNewAbstractMethodProblem]("org.apache.spark.ml.param.shared.HasValidationIndicatorCol.org$apache$spark$ml$param$shared$HasValidationIndicatorCol$_setter_$validationIndicatorCol_="),
-    ProblemFilters.exclude[InheritedNewAbstractMethodProblem]("org.apache.spark.ml.param.shared.HasValidationIndicatorCol.validationIndicatorCol")
+    ProblemFilters.exclude[InheritedNewAbstractMethodProblem]("org.apache.spark.ml.param.shared.HasValidationIndicatorCol.validationIndicatorCol"),
+
+    // [SPARK-23042] Use OneHotEncoderModel to encode labels in MultilayerPerceptronClassifier
+    ProblemFilters.exclude[MissingClassProblem]("org.apache.spark.ml.classification.LabelConverter")
   )
 
   // Exclude rules for 2.3.x