Spark机器学习中ml和mllib中矩阵、向量

 import org.apache.spark.ml.linalg.{DenseVector => MLDenseVector}
  val mlDv = new MLDenseVector(Array[Double](1, 1, 1, 1, 1))

  println(mlDv.argmax)

  //压缩矩阵，底层根据0的个数进行判断是稀疏存储还是稠密存储。稀疏存储就是存储非0的元素值以及索引以及向量的大小（也就是三维）
  println(mlDv.compressed)
  val copy = mlDv.copy //深拷贝

  copy.foreachActive {
    (x, y) =>
      println("index = " + x + " ,  value = " + y)
  }

  //Number of active entries. An "active entry" is an element which is explicitly(明确地) stored,
  // regardless of its value. Note that inactive entries have value 0.
  println(copy.numActives)
  println(copy.numNonzeros)
  println(copy.size)
  println(copy.values)
  println(copy.toSparse)

 import org.apache.spark.ml.linalg.{SparseVector => MLSparseVector}
  val mlDv = new MLDenseVector(Array[Double](1, 0, 0, 0, 0))
  println(mlDv.toSparse)  //(5,[0],[1.0])
  //SparseVector构造器：向量维度，非零索引，非零索引对应的值
  val mlSv = new MLSparseVector(5, Array[Int](0, 3), Array[Double](1, 2))
  println(mlSv) //(5,[0,3],[1.0,2.0])
  println(mlSv.toDense) //[1.0,0.0,0.0,2.0,0.0]
  println(mlSv.indices.toBuffer)//返回稀疏向量的索引

  //稀疏矩阵
  import org.apache.spark.mllib.linalg.{DenseVector => MLLIBDenseVector}
  val mlDv = new MLLIBDenseVector(Array[Double](1, 0, 0, 0, 0))
  mlDv.asML //直接转成spark ml的向量

import org.apache.spark.ml.linalg.{DenseMatrix => MLDenseMatrix}
  import org.apache.spark.ml.linalg.{SparseMatrix => MLSparseMatrix}

  // 默认以列为主的稠密矩阵。
  val notTranspose = new MLDenseMatrix(3, 2, Array[Double](1, 3, 5, 2, 4, 6))
  // 第三个参数为是否允许转至，默认不允许，如果允许则按行存储
  val mlDMtx = new MLDenseMatrix(3, 2, Array[Double](1, 2, 3, 4, 5, 6), true)

  println(notTranspose)

  println("-------------------------------------------------")
  println(notTranspose.isTransposed)
  println(notTranspose.transpose)
  println(mlDMtx.isTransposed)
  println("-------------------------------------------------")
  println(mlDMtx)
  println(mlDMtx.compressed)
  println("-------------------------------------------------")
  //转为按照列存储的稠密矩阵
  println(mlDMtx.toDenseColMajor)

  //转为按照行存储的稠密矩阵
  println(notTranspose.toDenseRowMajor)