ODIn/spark-instrumented-optimizer
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity

[SPARK-10875] [MLLIB] Computed covariance matrix should be symmetric

Browse source 
Compute upper triangular values of the covariance matrix, then copy to lower triangular values.

Author: Nick Pritchard <nicholas.pritchard@falkonry.com>

Closes #8940 from pnpritchard/SPARK-10875.
This commit is contained in:
Nick Pritchard 2015-10-08 22:22:20 -07:00 committed by Xiangrui Meng
parent 5410747a84
commit 5994cfe812
2 changed files with 22 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

6
mllib/src/main/scala/org/apache/spark/mllib/linalg/distributed/RowMatrix.scala
View file

@ -357,9 +357,11 @@ class RowMatrix @Since("1.0.0") (
var alpha = 0.0
while (i < n) {
alpha = m / m1 * mean(i)
j = 0
j = i
while (j < n) {
G(i, j) = G(i, j) / m1 - alpha * mean(j)
val Gij = G(i, j) / m1 - alpha * mean(j)
G(i, j) = Gij
G(j, i) = Gij
j += 1
}
i += 1

18
mllib/src/test/scala/org/apache/spark/mllib/linalg/distributed/RowMatrixSuite.scala
View file

@ -24,6 +24,7 @@ import breeze.linalg.{DenseVector => BDV, DenseMatrix => BDM, norm => brzNorm, s
import org.apache.spark.SparkFunSuite
import org.apache.spark.mllib.linalg.{Matrices, Vectors, Vector}
import org.apache.spark.mllib.random.RandomRDDs
import org.apache.spark.mllib.util.{LocalClusterSparkContext, MLlibTestSparkContext}
class RowMatrixSuite extends SparkFunSuite with MLlibTestSparkContext {
@ -255,6 +256,23 @@ class RowMatrixSuite extends SparkFunSuite with MLlibTestSparkContext {
assert(closeToZero(abs(expected.r) - abs(rOnly.R.toBreeze.asInstanceOf[BDM[Double]])))
}
}
test("compute covariance") {
for (mat <- Seq(denseMat, sparseMat)) {
val result = mat.computeCovariance()
val expected = breeze.linalg.cov(mat.toBreeze())
assert(closeToZero(abs(expected) - abs(result.toBreeze.asInstanceOf[BDM[Double]])))
}
}
test("covariance matrix is symmetric (SPARK-10875)") {
val rdd = RandomRDDs.normalVectorRDD(sc, 100, 10, 0, 0)
val matrix = new RowMatrix(rdd)
val cov = matrix.computeCovariance()
for (i <- 0 until cov.numRows; j <- 0 until i) {
assert(cov(i, j) === cov(j, i))
}
}
}
class RowMatrixClusterSuite extends SparkFunSuite with LocalClusterSparkContext {