Numpy basic operations

Numpy basic operations

One, build an array

Note: numpy can build arrays with lists and tuples

1. List build array

import numpy as np
data1 = [5, 7, 9, 20]    #列表
arr1 = np.array(data1)
arr1

2. Tuples to build an array

data2 = (5, 7, 9, 20)     #元组
arr2 = np.array(data2)
arr2

3. Build a multidimensional array

data3 = [[1, 2, 3, 4],[5, 6, 7, 8]]   #多维数组
arr3 = np.array(data3)
arr3

4. Array ranks and element types

arr3.shape
#数组的形状（行和列）
arr3.dtype
#数组元素类型

5. All 1 array and all 0 array

np.zeros(8)
#生成8个元素全为1的数组
np.zeros((3,4))
#生成3行4列的全1数组
np.ones(4)
#生成全0数组
np.ones((4,6))
#生产4行6列的全0数组
np.empty((2, 2, 2))
#生成随机的2行2列的2维数据

6.arange generates an array

np.arange(10)
#生成从1到10（不包括10）的数组

7. Change the array to all 1s

arr5 = np.ones_like(arr3)
#将arr3的数据全部改为1
arr5

8. The astype() function changes the data type of a column in the DataFrame

注：
int32 --> float64
float64 --> int32
string_ --> float64

arr3 = arr1.astype('string_')
#改arr1的数据类型为字符串数组

2. Data reshaping and merging

1. Convert the array shape

Note: When the reshape parameter is -1, the number of new rows (columns) is automatically calculated based on the original array

arr = np.arange(9)
arr.reshape((3,3))
#将数组转换为3行3列的数组

2. Convert a multi-dimensional array to a 1-dimensional array

Note:
ravel(): If not necessary, a copy of the source data will not be generated.
flatten(): Return a copy of the source data.
squeeze(): Only the dimension of dimension 1 can be reduced

arr.ravel()
#转换为1维数组
arr.flatten()
#返回源数据的副本

3. Concatenate function to concatenate arrays

np.concatenate([arr1, arr2],axis=0)
#连接这2个数组，按行连接
np.concatenate([arr1, arr2],axis=1)
#按列连接

4. The vstack function stacks the array vertically (in line order)

np.vstack((arr1,arr2))
#按行顺序把这2个数组连接起来

5.hstack level (in column order) to stack the array

np.hstack((arr1, arr2))
#按列顺序把2个数组连接起来

6.split（(ary, indices_or_sections, axis=0）

(1)ary: the array to be divided
(2) indicators_or_sections: if it is an integer, the number is divided equally, if it is an array, it is the position of the division along the axis (opening on the left and closed on the right)
(3)axis ：The dimension along which to perform the tangential direction, the default is 0, and the horizontal division is performed. When it is 1, split longitudinally

np.split(arr,[2, 4])
#将数组以2个数组平均切分，并横向切分（即按列切分）

7. Array transpose, transformation (transpose, swapaxes)

arr.transpose((1,0))
#行和列互换位置
arr.T
#数组转置
arr.swapaxes(1, 2)
#将第2轴和第3轴数据交换

Three, random number

arr = np.random.randint(100, 200, size=(5, 4))
#返回100到200之间随机的整数，5行4列的数组
arr = np.random.randn(2, 3, 5)
#返回2行5列的3维数组，符合标准正态分布
arr = np.random.normal(4, 5, size=(3,5))
#返回3行5列，μ = 4,σ = 5的正态分布样本
np.random.permutation(arr)
#返回一个数组的随机排序（返回一个新的，不改变原数组）
np.random.shuffle(arr)
#打乱数组的排列顺序（改变原数组）

Fourth, index and slice

numpy slice explanation

Note:
This is a numpy slicing operation, the general structure is num[a:b,c:d], the comma is used as the separator during analysis,
and the subscript range of the num line to be taken before the comma (a to b-1) , After the comma is the subscript range of the num column to be taken (c to d-1); the
front is the row index, and the back is the column index.
If it is this num[:b,c:d], the value of a is not specified, then a is the minimum value of 0;
if it is this num[a:,c:d], the value of b is not specified, then b is Maximum value; the same applies to the case of c and d.

1. Index

arr[3]
#索引第4个数据
arr[-1]
#索引最后一个数据
arr[2] = 123
#改第3个数据为123
arr1 = arr[-3:-1]
#选取数组的后2个数据（除过最后一个数据）到新变量
arr1[:] = 77
#将新变量所有数据改为77（***此时原数组的值也改变***）
arr1 = arr[1].copy()
#生成新变量的副本（***此时再改变变量，不会改变原数组的值***）
arr[0][3]
#对多维数组可用二次索引（选取第一行数组的第4个数据）
arr[2,3]
#对多维数组多次索引，与上作用相同（选取第3行数据的第4个数据）

2. Index column

arr[:, 1]
#索取第2列
arr[:, 1:2]
#索取第1列（不包括第2列）
arr[2:, 1:]
#索取第3行后，第2列后的数组

3. Judgment Index

fruits == 'pear'
#判定fruits数组的每个数据是否为pear，是就为True，否就为False
datas[fruits == 'pear']
#按True的位置索引datas数组中的数组
datas[(fruits == 'apple') | (fruits == 'banana')]
#并集，选取这2个为True的数组数据
arr[[1, 3, 2]]
#索引第2行，第4行，第3行的数据
arr[[3, 2]][:, [2, 1]]
arr[np.ix_([3,2],[2, 1])]
#索引为[3,2],[3,1]
       [2,2],[2,1]

Explanation of where function

ix_ function explanation

Five, other operations

Explanation of cumsum function

Unique function explanation

in1d function explanation

Explanation of savetxt function

Explanation of loadtxt function

np.modf(arr)
#返回数组的整数和小数部分
arr.sum()
#返回数组所有数据之和
arr.mean()
#返回平均值
arr.std()
#返回总体标准偏差
arr.cumsum(0)
#返回按指定轴累加后的元素
arr.cumprod(1)
#累乘函数
arr.sort()
#升序排列
np.unique(fruits)
#去除fruits数组中重复的数据，并按从大到小的顺序排列输出
np.in1d(arr, [2,7])
#若数据为2和7，返回True，反之则False
np.savetxt('ch2ex1.csv', arr, fmt='%d',delimiter=',')
#将arr数组保存为csv文件
arr = np.loadtxt('ch2ex1.csv', delimiter=',')
#读取csv文件
np.dot(arr1,arr2)
#返回这2个数组的点积
from numpy.linalg import det
det(arr)
#计算数组的行列式

Six, image processing

from PIL import Image
import numpy as np
im = np.array(Image.open('C:/Users/LP/Desktop/2.76.jpeg'))
#加载图片
print(im.shape,im.dtype)
#输出图片尺寸和格式
b = [255, 255, 255] - im    #数组运算
new_im = Image.fromarray(b.astype('uint8'))  
new_im.save('C:/Users/LP/Desktop/2.79.jpeg')  
#保存为新图像