[SPARK-21681][ML] fix bug of MLOR do not work correctly when featureStd contains zero (backport PR for 2.2)

## What changes were proposed in this pull request?

This is backport PR of apache#18896

fix bug of MLOR do not work correctly when featureStd contains zero

We can reproduce the bug through such dataset (features including zero variance), will generate wrong result (all coefficients becomes 0)
```
    val multinomialDatasetWithZeroVar = {
      val nPoints = 100
      val coefficients = Array(
        -0.57997, 0.912083, -0.371077,
        -0.16624, -0.84355, -0.048509)

      val xMean = Array(5.843, 3.0)
      val xVariance = Array(0.6856, 0.0)  // including zero variance

      val testData = generateMultinomialLogisticInput(
        coefficients, xMean, xVariance, addIntercept = true, nPoints, seed)

      val df = sc.parallelize(testData, 4).toDF().withColumn("weight", lit(1.0))
      df.cache()
      df
    }
```
## How was this patch tested?

testcase added.

Author: WeichenXu <[email protected]>

Closes apache#19026 from WeichenXu123/fix_mlor_zero_var_bug_2_2.

Author: WeichenXu123

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

@@ -1727,11 +1727,13 @@ private class LogisticAggregator(
 
     val margins = new Array[Double](numClasses)
     features.foreachActive { (index, value) =>
-      val stdValue = value / localFeaturesStd(index)
-      var j = 0
-      while (j < numClasses) {
-        margins(j) += localCoefficients(index * numClasses + j) * stdValue
-        j += 1
+      if (localFeaturesStd(index) != 0.0 && value != 0.0) {
+        val stdValue = value / localFeaturesStd(index)
+        var j = 0
+        while (j < numClasses) {
+          margins(j) += localCoefficients(index * numClasses + j) * stdValue
+          j += 1
+        }
       }
     }
     var i = 0

mllib/src/test/scala/org/apache/spark/ml/classification/LogisticRegressionSuite.scala

@@ -45,6 +45,7 @@ class LogisticRegressionSuite
   @transient var smallMultinomialDataset: Dataset[_] = _
   @transient var binaryDataset: Dataset[_] = _
   @transient var multinomialDataset: Dataset[_] = _
+  @transient var multinomialDatasetWithZeroVar: Dataset[_] = _
   private val eps: Double = 1e-5
 
   override def beforeAll(): Unit = {
@@ -98,6 +99,23 @@ class LogisticRegressionSuite
       df.cache()
       df
     }
+
+    multinomialDatasetWithZeroVar = {
+      val nPoints = 100
+      val coefficients = Array(
+        -0.57997, 0.912083, -0.371077,
+        -0.16624, -0.84355, -0.048509)
+
+      val xMean = Array(5.843, 3.0)
+      val xVariance = Array(0.6856, 0.0)
+
+      val testData = generateMultinomialLogisticInput(
+        coefficients, xMean, xVariance, addIntercept = true, nPoints, seed)
+
+      val df = sc.parallelize(testData, 4).toDF().withColumn("weight", lit(1.0))
+      df.cache()
+      df
+    }
   }
 
   /**
@@ -111,6 +129,11 @@ class LogisticRegressionSuite
     multinomialDataset.rdd.map { case Row(label: Double, features: Vector, weight: Double) =>
       label + "," + weight + "," + features.toArray.mkString(",")
     }.repartition(1).saveAsTextFile("target/tmp/LogisticRegressionSuite/multinomialDataset")
+    multinomialDatasetWithZeroVar.rdd.map {
+      case Row(label: Double, features: Vector, weight: Double) =>
+        label + "," + weight + "," + features.toArray.mkString(",")
+    }.repartition(1)
+      .saveAsTextFile("target/tmp/LogisticRegressionSuite/multinomialDatasetWithZeroVar")
   }
 
   test("params") {
@@ -1391,6 +1414,58 @@ class LogisticRegressionSuite
     assert(model2.interceptVector.toArray.sum ~== 0.0 absTol eps)
   }
 
+  test("multinomial logistic regression with zero variance (SPARK-21681)") {
+    val sqlContext = multinomialDatasetWithZeroVar.sqlContext
+    import sqlContext.implicits._
+    val mlr = new LogisticRegression().setFamily("multinomial").setFitIntercept(true)
+      .setElasticNetParam(0.0).setRegParam(0.0).setStandardization(true).setWeightCol("weight")
+
+    val model = mlr.fit(multinomialDatasetWithZeroVar)
+
+    /*
+     Use the following R code to load the data and train the model using glmnet package.
+     library("glmnet")
+     data <- read.csv("path", header=FALSE)
+     label = as.factor(data$V1)
+     w = data$V2
+     features = as.matrix(data.frame(data$V3, data$V4))
+     coefficients = coef(glmnet(features, label, weights=w, family="multinomial",
+     alpha = 0, lambda = 0))
+     coefficients
+     $`0`
+     3 x 1 sparse Matrix of class "dgCMatrix"
+                    s0
+             0.2658824
+     data.V3 0.1881871
+     data.V4 .
+     $`1`
+     3 x 1 sparse Matrix of class "dgCMatrix"
+                      s0
+              0.53604701
+     data.V3 -0.02412645
+     data.V4  .
+     $`2`
+     3 x 1 sparse Matrix of class "dgCMatrix"
+                     s0
+             -0.8019294
+     data.V3 -0.1640607
+     data.V4  .
+    */
+
+    val coefficientsR = new DenseMatrix(3, 2, Array(
+      0.1881871, 0.0,
+      -0.02412645, 0.0,
+      -0.1640607, 0.0), isTransposed = true)
+    val interceptsR = Vectors.dense(0.2658824, 0.53604701, -0.8019294)
+
+    model.coefficientMatrix.colIter.foreach(v => assert(v.toArray.sum ~== 0.0 absTol eps))
+
+    assert(model.coefficientMatrix ~== coefficientsR relTol 0.05)
+    assert(model.coefficientMatrix.toArray.sum ~== 0.0 absTol eps)
+    assert(model.interceptVector ~== interceptsR relTol 0.05)
+    assert(model.interceptVector.toArray.sum ~== 0.0 absTol eps)
+  }
+
   test("multinomial logistic regression with intercept without regularization with bound") {
     // Bound constrained optimization with bound on one side.
     val lowerBoundsOnCoefficients = Matrices.dense(3, 4, Array.fill(12)(1.0))