CV&NLP Basics 1 Numpy Learning

Numpy array processing library

Import the numpy library and alias it as np

import numpy as np

See what is provided in the numpy library

dir(np)或help(np)

Supplementary description of a function of jupyer

Put the cursor on the function and press shift+tabto call out the function description:

Use numpy's random function to randomly generate arrays

numpy.random.rand()

numpy.random.rand(d0,d1,…,dn)

• The rand function generates data between [0,1) according to the given dimension, including 0 and excluding 1
• dn indicates dimension
• The return value is an array of the specified dimension

example:

numpy.random.randn() - important

numpy.random.randn(d0,d1,…,dn)

• The randn function returns a sample or set of samples, with a standard normal distribution.

• dn indicates dimension

• The return value is an array of the specified dimension

• .shapeThe construction of the resulting array can be viewed via

• By .ndimlooking at the dimension of the array (also can be understood as the number of "axes")

• By .sizelooking at the number of elements contained in the array
  

• .argmax: Returns the index (subscript) of the maximum value for the specified axis. In a two-dimensional array, the parameter "0" represents the column, and the parameter "1" represents the row. After
  
  explaining why it is used .argmax(axis=0), the output array([1, 1, 0]) ?
  .argmaxreturns the maximum value of the specified axis 索引(下标). 二维数组中，参数“0”代表列，参数“1”代表行。
  Use .argmax(0)fetch 列的最大值的索引:
  The first column of the generated three-row three-column array arr3 is: 0.78650488, 1.30572907, -0.49828544, among which 1.30572907 is the largest, and its index is 1.
  The second column of the generated three-row three-column array arr3 is: 0.38190384, 1.60970205, -2.00096618, among which 1.60970205 is the largest, and its index is 1.
  The third column of the generated three-row three-column array arr3 is: 1.145984, -0.2094228, -0.85838297, among which 1.145984 is the largest, and its index is 0.

In the same way, use .argmax(axis=1)fetch 行的最大值的索引:

numpy.random.randint()

numpy.random.randint(low, high=None, size=None, dtype=numpy.int32)

• Return a random integer , the range is [low, high), including low, excluding high
• Parameters: low is the minimum value, high is the maximum value, size is the size of the array dimension, dtype is the data type, and the default data type is np.int
• When high is not filled in, the default range of random number generation is [0, low)

Generate a floating point number between [0,1)

• numpy.random.random_sample(size=None)
• numpy.random.random(size=None)
• numpy.random.ranf(size=None)
• numpy.random.sample(size=None)

numpy.random.choice()

numpy.random.choice(a, size=None, replace=True, p=None)

• Generate random numbers from a given 1D array
• Parameters: a is a one-dimensional array similar to data or an integer; size is the dimension of the array; p is the probability of the data in the array appearing
• When a is an integer, the corresponding one-dimensional array is np.arange(a)
  
• The length of parameter p needs to be consistent with the length of parameter a;
• The parameter p is the probability, and the sum of the data in p should be 1
  

numpy.random.seed()

• The role of np.random.seed(): Make random data predictable.
• When we set the same seed, the random number generated each time is the same. If no seed is set, a different random number will be generated each time
  

numpy.zeros()

It is used to generate an array of all 0s, and the variable type can be specified with the dtype parameter

Example: Generate an array of all zeros with three rows and four columns, and the type is float32

array multiplication numpy.dot()

numpy.arange() - important

Use the arange function in the numpy library to create a range of values ​​and return an ndarray (n-dimensional array, n-dimensional array) object. The format of the function is as follows:

numpy.arange(start, stop, step, dtype)

Generate an ndarray according to the range specified by start and stop and the step size set by step.

array broadcast

Multiplying an array by 10 will multiply all the numbers in it by 10

Add a number to the array, each element will add this number

If arr2 is an array of 4 rows and 6 columns, and arr3 is a matrix of 1 row and 6 columns, arr2 + arr3 is each row of arr2 plus the corresponding element of arr3

Features of the array type

• has its own shape (shape)
• There are data types (dtype), such as: dtype=numpy.int32; dtype=numpy.float32
• You can modify the value of an element in an array

Modify array shape numpy.reshape() - very important

1. Preliminary work, first generate an array object:

arr5 = np.arange(0, 24, dtype=np.int32)
arr5

2. Convert the array to an array of three rows and eight columns:

arr5.reshape(3, 8)

3. Convert the array to an array of four rows and six columns:

arr5.reshape(4, 6)

4. Find the first element of the array whose shape has been modified by indexing (note the difference from list indexing, array indexing is a bit simpler and less parentheses):
   

5. Indexed elements in supplementary lists:
   

6. Modify elements in an array:
   
   注意，使用reshape函数时，不会创建副本。例如，这里修改了arr6的元素，但是也修改了arr5的元素，arr5和arr6指向同一个元素，没有创建副本，如下图:

7. 使用copy函数，创建副本

arr6 = arr5.copy()

Use the function list() to force an array to be converted to a list

Use the function list() to force an array of three rows and four columns to be converted to a list, where each element is still an array type:

Use .ndim to view the dimensions of an array

It can also be understood as the number of axes in the array.
Example: arr6 is an array of three rows and four columns

Use .size to view the number of elements in the array

Type conversion (emphasis)

Use numpy.array() to convert a list to an array

Pay attention to the comments, involving the use of type() function and dtype attribute

numpy.astype()

By looking at the function description, we can see that astype() can create a copy of the original array and change the type of elements in the array

Example:

After using the astype() function, check whether it is a copy of arr7 through the elements in arr8:

array slice

数组切片是原始数据的视图，也就是不会创造副本。注意这与列表切片的区别。

High-dimensional array slices:
Example 1:

Example 2: