numpy3数组操作

数组操作

更改形状

numpy.ndarray.shape表示数组的维度，返回一个元组，这个元组的长度就是维度的数目，即 ndim 属性(秩)。
通过修改 shape 属性来改变数组的形状。

import numpy as np

x = np.array([1, 2, 9, 4, 5, 6, 7, 8])
print(x.shape)  # (8,)
x.shape = [2, 4]#将x矩阵改为两行三列
print(x)
# [[1 2 9 4]
#  [5 6 7 8]]

x = np.array([[11, 12, 13, 14, 15],
              [16, 17, 18, 19, 20],
              [21, 22, 23, 24, 25],
              [26, 27, 28, 29, 30],
              [31, 32, 33, 34, 35]])
y = x.flat
print(y)
# <numpy.flatiter object at 0x0000020F9BA10C60>
for i in y:
    print(i, end=' ')#输出y的每个值并用空格分开
# 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
y[3] = 0#将y中第四个值替换为0
print(end='\n')#换行，此处去掉无影响
print(x)
# [[11 12 13  0 15]
#  [16 17 18 19 20]
#  [21 22 23 24 25]
#  [26 27 28 29 30]
#  [31 32 33 34 35]]

flatten()函数返回的是拷贝。

import numpy as np
x = np.array([[11, 12, 13, 14, 15],
              [16, 17, 18, 19, 20],
              [21, 22, 23, 24, 25],
              [26, 27, 28, 29, 30],
              [31, 32, 33, 34, 35]])
y = x.flatten()
print(y)
# [11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
#  35]
y[3] = 0
print(y)#j将第四个值替换
#[11 12 13  0 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
 35
print(x)#由于flatten返回的是拷贝之后的，所以不改变x 的数据
# [[11 12 13 14 15]
#  [16 17 18 19 20]
#  [21 22 23 24 25]
#  [26 27 28 29 30]
#  [31 32 33 34 35]]

order：‘C’ – 按行，‘F’ – 按列，‘A’ – 原顺序，‘k’ – 元素在内存中的出现顺序

y = x.flatten(order='F')
print(y)
# [11 16 21 26 31 12 17 22 27 32 13 18 23 28 33 14 19 24 29 34 15 20 25 30 35]

ravel()返回的是视图。

import numpy as np

x = np.array([[11, 12, 13, 14, 15],
              [16, 17, 18, 19, 20],
              [21, 22, 23, 24, 25],
              [26, 27, 28, 29, 30],
              [31, 32, 33, 34, 35]])
y = np.ravel(x)
print(y)
# [11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
#  35]

y[3] = 0
print(x)
# [[11 12 13  0 15]
#  [16 17 18 19 20]
#  [21 22 23 24 25]
#  [26 27 28 29 30]
#  [31 32 33 34 35]]

加入order=F 就是拷贝

x = np.array([[11, 12, 13, 14, 15],
              [16, 17, 18, 19, 20],
              [21, 22, 23, 24, 25],
              [26, 27, 28, 29, 30],
              [31, 32, 33, 34, 35]])

y = np.ravel(x, order='F')
print(y)
# [11 16 21 26 31 12 17 22 27 32 13 18 23 28 33 14 19 24 29 34 15 20 25 30
#  35]

y[3] = 0
print(x)
# [[11 12 13 14 15]
#  [16 17 18 19 20]
#  [21 22 23 24 25]
#  [26 27 28 29 30]
#  [31 32 33 34 35]]

reshape()函数当参数newshape = [rows,-1]时，将根据行数自动确定列数。

x = np.arange(12)
y = np.reshape(x, [3, 4])
print(y.dtype)  # int32
print(y)
# [[ 0  1  2  3]
#  [ 4  5  6  7]
#  [ 8  9 10 11]]

y = np.reshape(x, [3, -1])
print(y)
# [[ 0  1  2  3]
#  [ 4  5  6  7]
#  [ 8  9 10 11]]

y = np.reshape(x,[-1,3])
print(y)
# [[ 0  1  2]
#  [ 3  4  5]
#  [ 6  7  8]
#  [ 9 10 11]]

y[0, 1] = 10
print(x)
# [ 0 10  2  3  4  5  6  7  8  9 10 11]（改变x去reshape后y中的值，x对应元素也改变）

reshape()函数当参数newshape = -1时，表示将数组降为一维。

x = np.random.randint(12, size=[2, 2, 3])
print(x)
# [[[11  9  1]
#   [ 1 10  3]]
# 
#  [[ 0  6  1]
#   [ 4 11  3]]]
y = np.reshape(x, -1)
print(y)
# [11  9  1  1 10  3  0  6  1  4 11  3]

数组转置

x = np.random.rand(5, 5) * 10#从0-1中随机取一个5*5矩阵，每个量乘10
x = np.around(x, 2) #保留两位小数？？?
print(x)
y = x.T#转置
y = np.transpose(x)#转置

很多工具包在进行计算时都会先判断输入数据的维度是否满足要求，如果输入数据达不到指定的维度时，可以使用newaxis参数来增加一个维度。

import numpy as np

x = np.array([1, 2, 9, 4, 5, 6, 7, 8])
print(x.shape)  # (8,)
print(x)  # [1 2 9 4 5 6 7 8]

y = x[np.newaxis, :]
print(y.shape)  # (1, 8)
print(y)  # [[1 2 9 4 5 6 7 8]]

y = x[:, np.newaxis]
print(y.shape)  # (8, 1)
print(y)

参考链接.