元素智能乘积

  ElementwiseProduct对每一个输入向量乘以一个给定的“权重”向量。换句话说，就是通过一个乘子对数据集的每一列进行缩放。这个转换可以表示为如下的形式：

6.1

  下面是一个使用的实例。

  1. import org.apache.spark.SparkContext._
  2. import org.apache.spark.mllib.feature.ElementwiseProduct
  3. import org.apache.spark.mllib.linalg.Vectors
  4. // Create some vector data; also works for sparse vectors
  5. val data = sc.parallelize(Array(Vectors.dense(1.0, 2.0, 3.0), Vectors.dense(4.0, 5.0, 6.0)))
  6. val transformingVector = Vectors.dense(0.0, 1.0, 2.0)
  7. val transformer = new ElementwiseProduct(transformingVector)
  9. // Batch transform and per-row transform give the same results:
  10. val transformedData = transformer.transform(data)
  11. val transformedData2 = data.map(x => transformer.transform(x))

  下面看transform的实现。

  1. override def transform(vector: Vector): Vector = {
  2.     vector match {
  3.       case dv: DenseVector =>
  4.         val values: Array[Double] = dv.values.clone()
  5.         val dim = scalingVec.size
  6.         var i = 0
  7.         while (i < dim) {
  8.           //相对应的值相乘
  9.           values(i) *= scalingVec(i)
  10.           i += 1
  11.         }
  12.         Vectors.dense(values)
  13.       case SparseVector(size, indices, vs) =>
  14.         val values = vs.clone()
  15.         val dim = values.length
  16.         var i = 0
  17.         while (i < dim) {
  18.           //相对应的值相乘
  19.           values(i) *= scalingVec(indices(i))
  20.           i += 1
  21.         }
  22.         Vectors.sparse(size, indices, values)
  23.       case v => throw new IllegalArgumentException("Does not support vector type " + v.getClass)
  24.     }
  25.   }