张量(Tensor),是Tensorflow里常用的一个数据结构

张量有很严谨的定义,可暂且把它简单理解为一个多维空间,一维张量就是向量,二维就是矩阵,三维直接就叫三维张量,Tensorflow就是围绕张量运算的一个库。——助教老师

LU

初识LU

Tensorflow的线性代数函数tf.linalg.lu

data = tf.constant([2, 1, 8, 7], dtype=tf.float32, shape=[2,2])
tf.linalg.lu(data)

output

Lu(lu=<tf.Tensor: shape=(2, 2), dtype=float32, numpy=
array([[ 8.  ,  7.  ],
       [ 0.25, -0.75]], dtype=float32)>, p=<tf.Tensor: shape=(2,), dtype=int32, numpy=array([1, 0])>)

tensorflow官方详解

tf.linalg.lu

Computes the LU decomposition of one or more square matrices.
The input is a tensor of shape […, M, M] whose inner-most 2 dimensions form square matrices.
The input has to be invertible.
The output consists of two tensors LU and P containing the LU decomposition of all input submatrices […, :, :]. LU encodes the lower triangular and upper triangular factors.
For each input submatrix of shape [M, M], L is a lower triangular matrix of shape [M, M] with unit diagonal whose entries correspond to the strictly lower triangular part of LU. U is a upper triangular matrix of shape [M, M] whose entries correspond to the upper triangular part, including the diagonal, of LU.
P represents a permutation matrix encoded as a list of indices each between 0 and M-1, inclusive. If P_mat denotes the permutation matrix corresponding to P, then the L, U and P satisfies P_mat * input = L * U.

输入限制

  • 最内部的两维需构成方阵(square matrix)
  • 可逆(invertible)

输出

  • LU矩阵:由L矩阵的下三角元素(不包括对角线),和U矩阵的上三角元素(包括对角线)组合而成
    • L(lower triangular),L矩阵的对角线是单位对角线(unit diagonal)
    • U(upper triangular)
  • P矩阵:置换矩阵(permutation matrix),取值为下标index
    • 置换矩阵参考
    • 置换矩阵由单位矩阵的行列交换而形成,用于构成矩阵的行列交换运算,可用于矩阵交换行后化简为上三角矩阵U

LU分解

LU分解可用于求解线性方程 Ax=b ,主要思路:

  • A

    =

    L

    U

    A=LU

    A=LU
    ,代入方程得

    L

    U

    x

    =

    b

    LUx=b

    LUx=b
    ,看作

    L

    y

    =

    b

    Ly=b

    Ly=b
    后,求得

    y

    y

    y
  • y

    =

    U

    x

    y=Ux

    y=Ux
    ,最终求得

    x

    x

    x

相当于表示一种把A化为上三角矩阵U的方式,即左乘

L

1

L^{-1}

L1

L

1

A

=

U

L^{-1}A=U

L1A=U
LU分解求解参考,解不一定唯一

SVD

初识SVD分解

data = tf.constant([[1, 2], [3, 4]], dtype=tf.float32)
tf.linalg.svd(data)

output

(<tf.Tensor: shape=(2,), dtype=float32, numpy=array([5.4649854 , 0.36596614], dtype=float32)>,
 <tf.Tensor: shape=(2, 2), dtype=float32, numpy=
 array([[-0.4045536 , -0.91451436],
        [-0.91451436,  0.40455353]], dtype=float32)>,
 <tf.Tensor: shape=(2, 2), dtype=float32, numpy=
 array([[-0.5760485 ,  0.81741554],
        [-0.81741554, -0.5760485 ]], dtype=float32)>)

奇异值分解(Singular Value Decomposition)

回顾特征值分解(EVD, eigen value decomposition):

  • A

    A

    A
    为实对称矩阵(方阵)
  • A

    =

    Q

    Λ

    Q

    T

    A=Q\Lambda Q^T

    A=QΛQT
    ,其中

    Λ

    \Lambda

    Λ
    是对角矩阵,

    Q

    Q

    Q
    是由特征向量构成的标准正交阵

EVD可用于降维,但对矩阵要求较高,SVD把EVD的概念拓展到普通矩阵上
SVD求解参考

(有待完善)


版权声明:本文为qq_39380838原创文章,遵循 CC 4.0 BY-SA 版权协议,转载请附上原文出处链接和本声明。
原文链接:https://blog.csdn.net/qq_39380838/article/details/123848880