From b692868af13a24b5e75a300d4a0ad9907b9a2512 Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Wed, 26 Mar 2014 15:12:19 -0700
Subject: [PATCH] Revert

---
 .../org/apache/spark/mllib/linalg/SVD.scala   | 14 -----------
 .../apache/spark/mllib/linalg/SVDSuite.scala  | 25 +++++++++----------
 2 files changed, 12 insertions(+), 27 deletions(-)

diff --git a/mllib/src/main/scala/org/apache/spark/mllib/linalg/SVD.scala b/mllib/src/main/scala/org/apache/spark/mllib/linalg/SVD.scala
index 9157eb9db7791..d5b7d9f855f03 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/linalg/SVD.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/linalg/SVD.scala
@@ -118,7 +118,6 @@ class SVD {
     val m = matrix.m
     val n = matrix.n
 
-    println("matrix "+matrix)
     if (m < n || m <= 0 || n <= 0) {
       throw new IllegalArgumentException("Expecting a tall and skinny matrix m=$m n=$n")
     }
@@ -130,7 +129,6 @@ class SVD {
     val rowIndices = matrix.rows.map(_.i)
 
     // compute SVD
-    println("Computing SVD")
     val (u, sigma, v) = denseSVD(matrix.rows.map(_.data))
 
     if (computeU) {
@@ -288,8 +286,6 @@ class SVD {
     // Compute A^T A, assuming rows are sparse enough to fit in memory
     val rows = data.map(entry =>
       (entry.i, (entry.j, entry.mval))).groupByKey()
-    println("rows "+rows)
-    rows.foreach{e => println("r "+e)}
     val emits = rows.flatMap {
       case (rowind, colsI) =>
         val cols = colsI.toList
@@ -301,26 +297,18 @@ class SVD {
             }
         }
     }.reduceByKey(_ + _)
-    println("emits ")
-    emits.foreach{e => println("e "+e)}
 
     // Construct jblas A^T A locally
     val ata = DoubleMatrix.zeros(n, n)
     for (entry <- emits.collect()) {
       ata.put(entry._1._1, entry._1._2, entry._2)
     }
-    println("ata "+ata)
 
     // Since A^T A is small, we can compute its SVD directly
     val svd = Singular.sparseSVD(ata)
-    println("init svd")
-    svd.foreach{x => println(x)}
-    println("done")
     val V = svd(0)
-    println("V" + V)
     // This will be updated to rcond
     val sigmas = MatrixFunctions.sqrt(svd(1)).toArray.filter(x => x > 1e-9)
-    println("Sigmas "+sigmas)
 
     if (sigmas.size < k) {
       throw new Exception("Not enough singular values to return k=" + k + " s=" + sigmas.size)
@@ -354,8 +342,6 @@ class SVD {
       // Multiply A by VS^-1
       val aCols = data.map(entry => (entry.j, (entry.i, entry.mval)))
       val bRows = vsirdd.map(entry => (entry._1._1, (entry._1._2, entry._2)))
-      aCols.foreach{x => println("a "+x)}
-      bRows.foreach{x => println("b "+x)}
       val retUdata = aCols.join(bRows).map {
         case (key, ((rowInd, rowVal), (colInd, colVal))) =>
           ((rowInd, colInd), rowVal * colVal)
diff --git a/mllib/src/test/scala/org/apache/spark/mllib/linalg/SVDSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/linalg/SVDSuite.scala
index 3a9811a72240b..20e2b0f84be06 100644
--- a/mllib/src/test/scala/org/apache/spark/mllib/linalg/SVDSuite.scala
+++ b/mllib/src/test/scala/org/apache/spark/mllib/linalg/SVDSuite.scala
@@ -87,17 +87,17 @@ class SVDSuite extends FunSuite with BeforeAndAfterAll {
     val retu = getDenseMatrix(u)
     val rets = getDenseMatrix(s)
     val retv = getDenseMatrix(v)
-
-
-    // check individual decomposition
+ 
+ 
+    // check individual decomposition  
     assertMatrixApproximatelyEquals(retu, svd(0))
     assertMatrixApproximatelyEquals(rets, DoubleMatrix.diag(svd(1)))
     assertMatrixApproximatelyEquals(retv, svd(2))
 
     // check multiplication guarantee
-    assertMatrixApproximatelyEquals(retu.mmul(rets).mmul(retv.transpose), denseA)
+    assertMatrixApproximatelyEquals(retu.mmul(rets).mmul(retv.transpose), denseA)  
   }
-/*
+
  test("dense full rank matrix svd") {
     val m = 10
     val n = 3
@@ -120,7 +120,7 @@ class SVDSuite extends FunSuite with BeforeAndAfterAll {
     val retv = new DoubleMatrix(v)
 
 
-    // check individual decomposition
+    // check individual decomposition  
     assertMatrixApproximatelyEquals(retu, svd(0))
     assertMatrixApproximatelyEquals(rets, DoubleMatrix.diag(svd(1)))
     assertMatrixApproximatelyEquals(retv, svd(2))
@@ -131,7 +131,7 @@ class SVDSuite extends FunSuite with BeforeAndAfterAll {
 
  test("rank one matrix svd") {
     val m = 10
-    val n = 3
+    val n = 3   
     val data = sc.makeRDD(Array.tabulate(m, n){ (a,b) =>
       MatrixEntry(a, b, 1.0) }.flatten )
     val k = 1
@@ -153,13 +153,13 @@ class SVDSuite extends FunSuite with BeforeAndAfterAll {
     val rets = getDenseMatrix(s)
     val retv = getDenseMatrix(v)
 
-    // check individual decomposition
+    // check individual decomposition  
     assertMatrixApproximatelyEquals(retu, svd(0).getColumn(0))
     assertMatrixApproximatelyEquals(rets, DoubleMatrix.diag(svd(1).getRow(0)))
     assertMatrixApproximatelyEquals(retv, svd(2).getColumn(0))
 
      // check multiplication guarantee
-    assertMatrixApproximatelyEquals(retu.mmul(rets).mmul(retv.transpose), denseA)
+    assertMatrixApproximatelyEquals(retu.mmul(rets).mmul(retv.transpose), denseA)  
   }
 
  test("truncated with k") {
@@ -168,7 +168,7 @@ class SVDSuite extends FunSuite with BeforeAndAfterAll {
     val data = sc.makeRDD(Array.tabulate(m,n){ (a, b) =>
       MatrixEntry(a, b, (a + 2).toDouble * (b + 1)/(1 + a + b)) }.flatten )
     val a = SparseMatrix(data, m, n)
-
+    
     val k = 1 // only one svalue above this
 
     val decomposed = new SVD().setK(k).compute(a)
@@ -185,11 +185,10 @@ class SVDSuite extends FunSuite with BeforeAndAfterAll {
     val retv = getDenseMatrix(v)
 
     assert(retrank == 1, "rank returned not one")
-
-    // check individual decomposition
+    
+    // check individual decomposition  
     assertMatrixApproximatelyEquals(retu, svd(0).getColumn(0))
     assertMatrixApproximatelyEquals(rets, DoubleMatrix.diag(svd(1).getRow(0)))
     assertMatrixApproximatelyEquals(retv, svd(2).getColumn(0))
   }
- */
 }