commit

mllib/src/main/scala/org/apache/spark/mllib/linalg/Vectors.scala

@@ -594,7 +594,7 @@ class DenseVector(val values: Array[Double]) extends Vector {
     new SparseVector(size, ii, vv)
   }
 
-  def argmax: Int = {
+  override def argmax: Int = {
     if (size == 0) {
       -1
     } else {
@@ -726,17 +726,42 @@ class SparseVector(
     } else {
 
       var maxIdx = 0
-      var maxValue = if(indices(0) != 0) 0 else values(0)
+      var maxValue = if(indices(0) != 0) 0.0 else values(0)
 
       foreachActive { (i, v) =>
-        if(v > maxValue){
+        if (v > maxValue) {
           maxIdx = i
           maxValue = v
         }
       }
+
+      // look for inactive values incase all active node values are negative
+      if(size != values.size && maxValue < 0){
+        maxIdx = calcInactiveIdx(indices(0))
+        maxValue = 0
+      }
       maxIdx
     }
   }
+
+  /**
+   * Calculates the first instance of an inactive node in a sparse vector and returns the Idx
+   * of the element.
+   * @param idx starting index of computation
+   * @return index of first inactive node or -1 if it cannot find one
+   */
+  private[SparseVector] def calcInactiveIdx(idx: Int): Int ={
+    if(idx < size){
+      if(!indices.contains(idx)){
+        idx
+      }else{
+        calcInactiveIdx(idx+1)
+      }
+    }else{
+      -1
+    }
+  }
+
 }
 
 object SparseVector {

mllib/src/test/scala/org/apache/spark/mllib/linalg/VectorsSuite.scala

@@ -91,10 +91,23 @@ class VectorsSuite extends FunSuite {
     val max = vec2.argmax
     assert(max === 3)
 
-    // check for case that sparse vector is created with only negative vaues {0.0,0.0,-1.0,0.0,-0.7}
-    val vec3 = Vectors.sparse(5,Array(2, 4),Array(-1.0,-.7))
+    val vec3 = Vectors.sparse(5,Array(2, 4),Array(1.0,-.7))
     val max2 = vec3.argmax
-    assert(max2 === 0)
+    assert(max2 === 2)
+
+    // check for case that sparse vector is created with only negative vaues {0.0, 0.0,-1.0, -0.7, 0.0}
+    val vec4 = Vectors.sparse(5,Array(2, 3),Array(-1.0,-.7))
+    val max3 = vec4.argmax
+    assert(max3 === 0)
+
+    // check for case that sparse vector is created with only negative vaues {-1.0, 0.0, -0.7, 0.0, 0.0}
+    val vec5 = Vectors.sparse(5,Array(0, 3),Array(-1.0,-.7))
+    val max4 = vec5.argmax
+    assert(max4 === 1)
+
+    val vec6 = Vectors.sparse(5,Array(0, 1, 2),Array(-1.0, -.025, -.7))
+    val max5 = vec6.argmax
+    assert(max5 === 3)
   }
 
   test("vector equals") {