[WIP] [SPARK-1328] Add vector statistics

As with the new vector system in MLlib, we find that it is good to add some new APIs to precess the `RDD[Vector]`. Beside, the former implementation of `computeStat` is not stable which could loss precision, and has the possibility to cause `Nan` in scientific computing, just as said in the [SPARK-1328](https://spark-project.atlassian.net/browse/SPARK-1328).

APIs contain:

* rowMeans(): RDD[Double]
* rowNorm2(): RDD[Double]
* rowSDs(): RDD[Double]
* colMeans(): Vector
* colMeans(size: Int): Vector
* colNorm2(): Vector
* colNorm2(size: Int): Vector
* colSDs(): Vector
* colSDs(size: Int): Vector
* maxOption((Vector, Vector) => Boolean): Option[Vector]
* minOption((Vector, Vector) => Boolean): Option[Vector]
* rowShrink(): RDD[Vector]
* colShrink(): RDD[Vector]

This is working in process now, and some more APIs will add to `LabeledPoint`. Moreover, the implicit declaration will move from `MLUtils` to `MLContext` later.

Author: Xusen Yin <[email protected]>
Author: Xiangrui Meng <[email protected]>

Closes #268 from yinxusen/vector-statistics and squashes the following commits:

d61363f [Xusen Yin] rebase to latest master
16ae684 [Xusen Yin] fix minor error and remove useless method
10cf5d3 [Xusen Yin] refine some return type
b064714 [Xusen Yin] remove computeStat in MLUtils
cbbefdb [Xiangrui Meng] update multivariate statistical summary interface and clean tests
4eaf28a [Xusen Yin] merge VectorRDDStatistics into RowMatrix
48ee053 [Xusen Yin] fix minor error
e624f93 [Xusen Yin] fix scala style error
1fba230 [Xusen Yin] merge while loop together
69e1f37 [Xusen Yin] remove lazy eval, and minor memory footprint
548e9de [Xusen Yin] minor revision
86522c4 [Xusen Yin] add comments on functions
dc77e38 [Xusen Yin] test sparse vector RDD
18cf072 [Xusen Yin] change def to lazy val to make sure that the computations in function be evaluated only once
f7a3ca2 [Xusen Yin] fix the corner case of maxmin
967d041 [Xusen Yin] full revision with Aggregator class
138300c [Xusen Yin] add new Aggregator class
1376ff4 [Xusen Yin] rename variables and adjust code
4a5c38d [Xusen Yin] add scala doc, refine code and comments
036b7a5 [Xusen Yin] fix the bug of Nan occur
f6e8e9a [Xusen Yin] add sparse vectors test
4cfbadf [Xusen Yin] fix bug of min max
4e4fbd1 [Xusen Yin] separate seqop and combop out as independent functions
a6d5a2e [Xusen Yin] rewrite for only computing non-zero elements
3980287 [Xusen Yin] rename variables
62a2c3e [Xusen Yin] use axpy and in-place if possible
9a75ebd [Xusen Yin] add case class to wrap return values
d816ac7 [Xusen Yin] remove useless APIs
c4651bb [Xusen Yin] remove row-wise APIs and refine code
1338ea1 [Xusen Yin] all-in-one version test passed
cc65810 [Xusen Yin] add parallel mean and variance
9af2e95 [Xusen Yin] refine the code style
ad6c82d [Xusen Yin] add shrink test
e09d5d2 [Xusen Yin] add scala docs and refine shrink method
8ef3377 [Xusen Yin] pass all tests
28cf060 [Xusen Yin] fix error of column means
54b19ab [Xusen Yin] add new API to shrink RDD[Vector]
8c6c0e1 [Xusen Yin] add basic statistics

mllib/src/main/scala/org/apache/spark/mllib/linalg/distributed/RowMatrix.scala

@@ -19,14 +19,146 @@ package org.apache.spark.mllib.linalg.distributed
 
 import java.util
 
-import breeze.linalg.{DenseMatrix => BDM, DenseVector => BDV, svd => brzSvd}
+import breeze.linalg.{Vector => BV, DenseMatrix => BDM, DenseVector => BDV, svd => brzSvd}
 import breeze.numerics.{sqrt => brzSqrt}
 import com.github.fommil.netlib.BLAS.{getInstance => blas}
 
 import org.apache.spark.annotation.Experimental
 import org.apache.spark.mllib.linalg._
 import org.apache.spark.rdd.RDD
 import org.apache.spark.Logging
+import org.apache.spark.mllib.stat.MultivariateStatisticalSummary
+
+/**
+ * Column statistics aggregator implementing
+ * [[org.apache.spark.mllib.stat.MultivariateStatisticalSummary]]
+ * together with add() and merge() function.
+ * A numerically stable algorithm is implemented to compute sample mean and variance:
+  *[[http://en.wikipedia.org/wiki/Algorithms_for_calculating_variance variance-wiki]].
+ * Zero elements (including explicit zero values) are skipped when calling add() and merge(),
+ * to have time complexity O(nnz) instead of O(n) for each column.
+ */
+private class ColumnStatisticsAggregator(private val n: Int)
+    extends MultivariateStatisticalSummary with Serializable {
+
+  private val currMean: BDV[Double] = BDV.zeros[Double](n)
+  private val currM2n: BDV[Double] = BDV.zeros[Double](n)
+  private var totalCnt = 0.0
+  private val nnz: BDV[Double] = BDV.zeros[Double](n)
+  private val currMax: BDV[Double] = BDV.fill(n)(Double.MinValue)
+  private val currMin: BDV[Double] = BDV.fill(n)(Double.MaxValue)
+
+  override def mean: Vector = {
+    val realMean = BDV.zeros[Double](n)
+    var i = 0
+    while (i < n) {
+      realMean(i) = currMean(i) * nnz(i) / totalCnt
+      i += 1
+    }
+    Vectors.fromBreeze(realMean)
+  }
+
+  override def variance: Vector = {
+    val realVariance = BDV.zeros[Double](n)
+
+    val denominator = totalCnt - 1.0
+
+    // Sample variance is computed, if the denominator is less than 0, the variance is just 0.
+    if (denominator > 0.0) {
+      val deltaMean = currMean
+      var i = 0
+      while (i < currM2n.size) {
+        realVariance(i) =
+          currM2n(i) + deltaMean(i) * deltaMean(i) * nnz(i) * (totalCnt - nnz(i)) / totalCnt
+        realVariance(i) /= denominator
+        i += 1
+      }
+    }
+
+    Vectors.fromBreeze(realVariance)
+  }
+
+  override def count: Long = totalCnt.toLong
+
+  override def numNonzeros: Vector = Vectors.fromBreeze(nnz)
+
+  override def max: Vector = {
+    var i = 0
+    while (i < n) {
+      if ((nnz(i) < totalCnt) && (currMax(i) < 0.0)) currMax(i) = 0.0
+      i += 1
+    }
+    Vectors.fromBreeze(currMax)
+  }
+
+  override def min: Vector = {
+    var i = 0
+    while (i < n) {
+      if ((nnz(i) < totalCnt) && (currMin(i) > 0.0)) currMin(i) = 0.0
+      i += 1
+    }
+    Vectors.fromBreeze(currMin)
+  }
+
+  /**
+   * Aggregates a row.
+   */
+  def add(currData: BV[Double]): this.type = {
+    currData.activeIterator.foreach {
+      case (_, 0.0) => // Skip explicit zero elements.
+      case (i, value) =>
+        if (currMax(i) < value) {
+          currMax(i) = value
+        }
+        if (currMin(i) > value) {
+          currMin(i) = value
+        }
+
+        val tmpPrevMean = currMean(i)
+        currMean(i) = (currMean(i) * nnz(i) + value) / (nnz(i) + 1.0)
+        currM2n(i) += (value - currMean(i)) * (value - tmpPrevMean)
+
+        nnz(i) += 1.0
+    }
+
+    totalCnt += 1.0
+    this
+  }
+
+  /**
+   * Merges another aggregator.
+   */
+  def merge(other: ColumnStatisticsAggregator): this.type = {
+    require(n == other.n, s"Dimensions mismatch. Expecting $n but got ${other.n}.")
+
+    totalCnt += other.totalCnt
+    val deltaMean = currMean - other.currMean
+
+    var i = 0
+    while (i < n) {
+      // merge mean together
+      if (other.currMean(i) != 0.0) {
+        currMean(i) = (currMean(i) * nnz(i) + other.currMean(i) * other.nnz(i)) /
+          (nnz(i) + other.nnz(i))
+      }
+      // merge m2n together
+      if (nnz(i) + other.nnz(i) != 0.0) {
+        currM2n(i) += other.currM2n(i) + deltaMean(i) * deltaMean(i) * nnz(i) * other.nnz(i) /
+          (nnz(i) + other.nnz(i))
+      }
+      if (currMax(i) < other.currMax(i)) {
+        currMax(i) = other.currMax(i)
+      }
+      if (currMin(i) > other.currMin(i)) {
+        currMin(i) = other.currMin(i)
+      }
+      i += 1
+    }
+
+    nnz += other.nnz
+    this
+  }
+}
 
 /**
  * :: Experimental ::
@@ -182,13 +314,7 @@ class RowMatrix(
       combOp = (s1: (Long, BDV[Double]), s2: (Long, BDV[Double])) => (s1._1 + s2._1, s1._2 += s2._2)
     )
 
-    // Update _m if it is not set, or verify its value.
-    if (nRows <= 0L) {
-      nRows = m
-    } else {
-      require(nRows == m,
-        s"The number of rows $m is different from what specified or previously computed: ${nRows}.")
-    }
+    updateNumRows(m)
 
     mean :/= m.toDouble
 
@@ -240,6 +366,19 @@ class RowMatrix(
     }
   }
 
+  /**
+   * Computes column-wise summary statistics.
+   */
+  def computeColumnSummaryStatistics(): MultivariateStatisticalSummary = {
+    val zeroValue = new ColumnStatisticsAggregator(numCols().toInt)
+    val summary = rows.map(_.toBreeze).aggregate[ColumnStatisticsAggregator](zeroValue)(
+      (aggregator, data) => aggregator.add(data),
+      (aggregator1, aggregator2) => aggregator1.merge(aggregator2)
+    )
+    updateNumRows(summary.count)
+    summary
+  }
+
   /**
    * Multiply this matrix by a local matrix on the right.
    *
@@ -276,6 +415,16 @@ class RowMatrix(
     }
     mat
   }
+
+  /** Updates or verfires the number of rows. */
+  private def updateNumRows(m: Long) {
+    if (nRows <= 0) {
+      nRows == m
+    } else {
+      require(nRows == m,
+        s"The number of rows $m is different from what specified or previously computed: ${nRows}.")
+    }
+  }
 }
 
 object RowMatrix {

mllib/src/main/scala/org/apache/spark/mllib/stat/MultivariateStatisticalSummary.scala

@@ -0,0 +1,56 @@
+/*
+ * 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
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.spark.mllib.stat
+
+import org.apache.spark.mllib.linalg.Vector
+
+/**
+ * Trait for multivariate statistical summary of a data matrix.
+ */
+trait MultivariateStatisticalSummary {
+
+  /**
+   * Sample mean vector.
+   */
+  def mean: Vector
+
+  /**
+   * Sample variance vector. Should return a zero vector if the sample size is 1.
+   */
+  def variance: Vector
+
+  /**
+   * Sample size.
+   */
+  def count: Long
+
+  /**
+   * Number of nonzero elements (including explicitly presented zero values) in each column.
+   */
+  def numNonzeros: Vector
+
+  /**
+   * Maximum value of each column.
+   */
+  def max: Vector
+
+  /**
+   * Minimum value of each column.
+   */
+  def min: Vector
+}

mllib/src/main/scala/org/apache/spark/mllib/util/MLUtils.scala

@@ -17,14 +17,13 @@
 
 package org.apache.spark.mllib.util
 
-import breeze.linalg.{Vector => BV, DenseVector => BDV, SparseVector => BSV,
-  squaredDistance => breezeSquaredDistance}
+import breeze.linalg.{Vector => BV, SparseVector => BSV, squaredDistance => breezeSquaredDistance}
 
 import org.apache.spark.annotation.Experimental
 import org.apache.spark.SparkContext
 import org.apache.spark.rdd.RDD
 import org.apache.spark.mllib.regression.LabeledPoint
-import org.apache.spark.mllib.linalg.{Vector, Vectors}
+import org.apache.spark.mllib.linalg.Vectors
 
 /**
  * Helper methods to load, save and pre-process data used in ML Lib.
@@ -158,58 +157,6 @@ object MLUtils {
     dataStr.saveAsTextFile(dir)
   }
 
-  /**
-   * Utility function to compute mean and standard deviation on a given dataset.
-   *
-   * @param data - input data set whose statistics are computed
-   * @param numFeatures - number of features
-   * @param numExamples - number of examples in input dataset
-   *
-   * @return (yMean, xColMean, xColSd) - Tuple consisting of
-   *     yMean - mean of the labels
-   *     xColMean - Row vector with mean for every column (or feature) of the input data
-   *     xColSd - Row vector standard deviation for every column (or feature) of the input data.
-   */
-  private[mllib] def computeStats(
-      data: RDD[LabeledPoint],
-      numFeatures: Int,
-      numExamples: Long): (Double, Vector, Vector) = {
-    val brzData = data.map { case LabeledPoint(label, features) =>
-      (label, features.toBreeze)
-    }
-    val aggStats = brzData.aggregate(
-      (0L, 0.0, BDV.zeros[Double](numFeatures), BDV.zeros[Double](numFeatures))
-    )(
-      seqOp = (c, v) => (c, v) match {
-        case ((n, sumLabel, sum, sumSq), (label, features)) =>
-          features.activeIterator.foreach { case (i, x) =>
-            sumSq(i) += x * x
-          }
-          (n + 1L, sumLabel + label, sum += features, sumSq)
-      },
-      combOp = (c1, c2) => (c1, c2) match {
-        case ((n1, sumLabel1, sum1, sumSq1), (n2, sumLabel2, sum2, sumSq2)) =>
-          (n1 + n2, sumLabel1 + sumLabel2, sum1 += sum2, sumSq1 += sumSq2)
-      }
-    )
-    val (nl, sumLabel, sum, sumSq) = aggStats
-
-    require(nl > 0, "Input data is empty.")
-    require(nl == numExamples)
-
-    val n = nl.toDouble
-    val yMean = sumLabel / n
-    val mean = sum / n
-    val std = new Array[Double](sum.length)
-    var i = 0
-    while (i < numFeatures) {
-      std(i) = sumSq(i) / n - mean(i) * mean(i)
-      i += 1
-    }
-
-    (yMean, Vectors.fromBreeze(mean), Vectors.dense(std))
-  }
-
   /**
    * Returns the squared Euclidean distance between two vectors. The following formula will be used
    * if it does not introduce too much numerical error:

mllib/src/test/scala/org/apache/spark/mllib/linalg/distributed/RowMatrixSuite.scala

@@ -170,4 +170,19 @@ class RowMatrixSuite extends FunSuite with LocalSparkContext {
       ))
     }
   }
+
+  test("compute column summary statistics") {
+    for (mat <- Seq(denseMat, sparseMat)) {
+      val summary = mat.computeColumnSummaryStatistics()
+      // Run twice to make sure no internal states are changed.
+      for (k <- 0 to 1) {
+        assert(summary.mean === Vectors.dense(4.5, 3.0, 4.0), "mean mismatch")
+        assert(summary.variance === Vectors.dense(15.0, 10.0, 10.0), "variance mismatch")
+        assert(summary.count === m, "count mismatch.")
+        assert(summary.numNonzeros === Vectors.dense(3.0, 3.0, 4.0), "nnz mismatch")
+        assert(summary.max === Vectors.dense(9.0, 7.0, 8.0), "max mismatch")
+        assert(summary.min === Vectors.dense(0.0, 0.0, 1.0), "column mismatch.")
+      }
+    }
+  }
 }

mllib/src/test/scala/org/apache/spark/mllib/util/MLUtilsSuite.scala

@@ -27,7 +27,6 @@ import com.google.common.base.Charsets
 import com.google.common.io.Files
 
 import org.apache.spark.mllib.linalg.Vectors
-import org.apache.spark.mllib.regression.LabeledPoint
 import org.apache.spark.mllib.util.MLUtils._
 
 class MLUtilsSuite extends FunSuite with LocalSparkContext {
@@ -56,18 +55,6 @@ class MLUtilsSuite extends FunSuite with LocalSparkContext {
     }
   }
 
-  test("compute stats") {
-    val data = Seq.fill(3)(Seq(
-      LabeledPoint(1.0, Vectors.dense(1.0, 2.0, 3.0)),
-      LabeledPoint(0.0, Vectors.dense(3.0, 4.0, 5.0))
-    )).flatten
-    val rdd = sc.parallelize(data, 2)
-    val (meanLabel, mean, std) = MLUtils.computeStats(rdd, 3, 6)
-    assert(meanLabel === 0.5)
-    assert(mean === Vectors.dense(2.0, 3.0, 4.0))
-    assert(std === Vectors.dense(1.0, 1.0, 1.0))
-  }
-
   test("loadLibSVMData") {
     val lines =
       """