spark 矩阵转置 java_sparkmllib矩阵向量

Spark MLlib底层的向量、矩阵运算使用了Breeze库，Breeze库提供了Vector/Matrix的实现以及相应计算的接口(Linalg)。但是在MLlib里面同时也提供了Vector和Linalg等的实现。 使用需导入：

import breeze.linalg._

import breeze.numerics._

1

2

Breeze创建函数

val m1 = DenseMatrix.zeros[Double](2,3)

1

DenseMatrix[Double] = 0.0 0.0 0.0 0.0 0.0 0.0

val v1 = DenseVector.zeros[Double](3)

1

DenseVector(0.0, 0.0, 0.0)

val v2 = DenseVector.ones[Double](3)

1

DenseVector(1.0, 1.0, 1.0)

val v3 =DenseVector.fill(3){5.0}

1

DenseVector(5.0, 5.0, 5.0)

val v4 =DenseVector.range(1,10,2)

1

DenseVector(1, 3, 5, 7, 9)

val m2 = DenseMatrix.eye[Double](3)

1

DenseMatrix[Double] = 1.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0

val v6 =diag(DenseVector(1.0,2.0,3.0))

1

DenseMatrix[Double] = 1.0 0.0 0.0 0.0 2.0 0.0 0.0 0.0 3.0

val v8 =DenseVector(1,2,3,4)

1

DenseVector(1, 2, 3, 4)

val v9 =DenseVector(1,2,3,4).t

1

Transpose(DenseVector(1, 2, 3, 4))

val v10 =DenseVector.tabulate(3){i => 2*i}

1

DenseVector(0, 2, 4)

val m4 = DenseMatrix.tabulate(3, 2){case (i, j) => i+j}

1

DenseMatrix[Int] = 0 1 1 2 2 3

val v11 = new DenseVector(Array(1, 2, 3, 4))

1

DenseVector(1, 2, 3, 4)

val m5 = new DenseMatrix(2, 3, Array(11, 12, 13, 21, 22, 23))

1

DenseMatrix[Int] = 11 13 22 12 21 23

val v12 =DenseVector.rand(4)

1

DenseVector(0.7517657487447951, 0.8171495400874123, 0.8923542318540489, 0.174311259949119)

val m6 =DenseMatrix.rand(2, 3)

1

DenseMatrix[Double] = 0.5349430131148125 0.8822136832272578 0.7946323804433382 0.41097756311601086 0.3181490074596882 0.34195102205697414

Breeze元素访问

val a =DenseVector(1,2,3,4,5,6,7,8,9,10)

1

DenseVector(1, 2, 3, 4, 5, 6, 7, 8, 9, 10)

a(1 to 4)

1

DenseVector(2, 3, 4, 5)

a(5 to 0 by -1)

1

DenseVector(6, 5, 4, 3, 2, 1)

a(1 to -1)

1

DenseVector(2, 3, 4, 5, 6, 7, 8, 9, 10)

a( -1 )

1

Int = 10

val m =DenseMatrix((1.0,2.0,3.0), (3.0,4.0,5.0))

1

DenseMatrix[Double] = 1.0 2.0 3.0 3.0 4.0 5.0

m(0,1)

1

Double = 2.0

m(::,1)

1

DenseVector(2.0, 4.0)

Breeze元素操作

val m =DenseMatrix((1.0,2.0,3.0), (3.0,4.0,5.0))

1

DenseMatrix[Double] = 1.0 2.0 3.0 3.0 4.0 5.0

m.reshape(3, 2) //从列开始计数

1

DenseMatrix[Double] = 1.0 4.0 3.0 3.0 2.0 5.0

m.toDenseVector

1

DenseVector(1.0, 3.0, 2.0, 4.0, 3.0, 5.0)

val m =DenseMatrix((1.0,2.0,3.0), (4.0,5.0,6.0) , (7.0,8.0,9.0))

1

DenseMatrix[Double] = 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0

lowerTriangular(m)

1

DenseMatrix[Double] = 1.0 0.0 0.0 4.0 5.0 0.0 7.0 8.0 9.0

upperTriangular(m)

1

DenseMatrix[Double] = 1.0 2.0 3.0 0.0 5.0 6.0 0.0 0.0 9.0

m.copy

1

linalg.DenseMatrix[Double] = 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0

diag(m)

1

DenseVector(1.0, 5.0, 9.0)

m(::, 2) :=5.0

1

DenseVector(5.0, 5.0, 5.0)

m

1

DenseMatrix[Double] = 1.0 2.0 5.0 4.0 5.0 5.0 7.0 8.0 5.0

m(1 to 2,1 to 2) := 5.0

1

DenseMatrix[Double] = 5.0 5.0 5.0 5.0

val a =DenseVector(1,2,3,4,5,6,7,8,9,10)

1

DenseVector(1, 2, 3, 4, 5, 6, 7, 8, 9, 10)

a(1 to 4) := 5

1

DenseVector(5, 5, 5, 5)

a(1 to 4) := DenseVector(1,2,3,4)

1

DenseVector(1, 2, 3, 4)

val a1 = DenseMatrix((1.0,2.0,3.0), (4.0,5.0,6.0))

val a2 = DenseMatrix((1.0,1.0,1.0), (2.0,2.0,2.0))

DenseMatrix.vertcat(a1,a2)

1

2

3

DenseMatrix[Double] = 1.0 2.0 3.0 4.0 5.0 6.0 1.0 1.0 1.0 2.0 2.0 2.0

DenseMatrix.horzcat(a1,a2)

1

DenseMatrix[Double] = 1.0 2.0 3.0 1.0 1.0 1.0 4.0 5.0 6.0 2.0 2.0 2.0

val b1 = DenseVector(1,2,3,4)

val b2 = DenseVector(1,1,1,1)

DenseVector.vertcat(b1,b2)

1

2

3

DenseVector(1, 2, 3, 4, 1, 1, 1, 1)

Breeze数值计算函数

val a = DenseMatrix((1.0,2.0,3.0), (4.0,5.0,6.0))

val b = DenseMatrix((1.0,1.0,1.0), (2.0,2.0,2.0))

a + b

1

2

3

DenseMatrix[Double] = 2.0 3.0 4.0 6.0 7.0 8.0

a :* b

1

DenseMatrix[Double] = 1.0 2.0 3.0 8.0 10.0 12.0

a :/ b

1

DenseMatrix[Double] = 1.0 2.0 3.0 2.0 2.5 3.0

a :< b

1

DenseMatrix[Boolean] = false false false false false false

a :==b

1

DenseMatrix[Boolean] = true false false false false false

a :+=1.0

1

DenseMatrix[Double] = 2.0 3.0 4.0 5.0 6.0 7.0

a :*=2.0

1

DenseMatrix[Double] = 4.0 6.0 8.0 10.0 12.0 14.0

max(a)

1

Double = 14.0

argmax(a)

1

(Int, Int) = (1,2)

DenseVector(1, 2, 3, 4) dot DenseVector(1, 1, 1, 1)//点积

1

Int = 10

Breeze求和函数

val a = DenseMatrix((1.0,2.0,3.0), (4.0,5.0,6.0) , (7.0,8.0,9.0))

sum(a)

1

2

3

Double = 45.0

sum(a, Axis._0)//每列求和

1

DenseMatrix[Double] = 12.0 15.0 18.0

sum(a, Axis._1)//按行求和

trace(a)//对角线求和 15

1

2

accumulate(DenseVector(1, 2, 3, 4)) //累计和 1+2 、1+2+3

1

DenseVector(1, 3, 6, 10)

Breeze布尔函数

val a = DenseVector(true, false, true)

val b = DenseVector(false, true, true)

a :& b

a :| b

!a

1

2

3

4

5

DenseVector(false, false, true)

val a = DenseVector(1.0, 0.0, -2.0)

any(a)//任一元素非0，true

all(a)//所有元素非0，false

1

2

3

Breeze线性代数函数

a \ b//线性求解

a.t//转置

det(a)//求特征值

inv(a)//求逆

pinv(a)//求伪逆

norm(a)//求范数

eigSym(a)//特征值和特征向量

val (er, ei, _) = eig(a) (实部与虚部分开)//特征值

eig(a)._3//特征向量

val svd.SVD(u,s,v) = svd(a)//奇异值分解

rank(a)//求矩阵的秩

a.length//矩阵长度

a.rows//矩阵行数

a.cols//矩阵列数

1

2

3

4

5

6

7

8

9

10

11

12

13

14

DenseMatrix((1.0,2.0,3.0), (4.0,5.0,6.0) , (7.0,8.0,9.0))

DenseMatrix((1.0,1.0,1.0), (1.0,1.0,1.0) , (1.0,1.0,1.0))

a \ b

a.t

1

2

3

4

DenseMatrix[Double] = 1.0 4.0 7.0 2.0 5.0 8.0 3.0 6.0 9.0

Breeze取整函数

round(a)//四舍五入

ceil(a)

floor(a)

signum(a)//符号函数

abs(a)

1

2

3

4

5

val a = DenseVector(1.2, 0.6, -2.3)

signum(a)

1

2

DenseVector(1.0, 1.0, -1.0)

Breeze其它函数

Breeze三角函数包括：

sin, sinh, asin, asinh

cos, cosh, acos, acosh

tan, tanh, atan, atanh

atan2

sinc(x) ，即sin(x)/x

sincpi(x) ，即 sinc(x * Pi)

1

2

3

4

5

6

Breeze对数和指数函数 Breeze对数和指数函数包括：

log, exp log10

log1p, expm1

sqrt, sbrt

pow

1

2

3

4

BLAS介绍(一个线性代数库)

BLAS按照功能被分为三个级别： Level 1：矢量-矢量运算，比如点积(ddot)，加法和数乘 (daxpy)， 绝对值的和(dasum)，等等； Level 2：矩阵-矢量运算，最重要的函数是一般的矩阵向量乘法(dgemv)； Level 3：矩阵-矩阵运算，最重要的函数是一般的矩阵乘法 (dgemm)； 每一种函数操作都区分不同数据类型(单精度、双精度、复数)