Introduction

From now on, I will explain the scientific calculation library used in machine learning over several articles. This time is Numpy edition. We will focus on the functions used in machine learning, not the whole.

Also, since I will write the code in Jupyter Notebook, Note that it behaves differently from the command prompt.

Version confirmation
Numpy Basics
Finally
References

Version confirmation

`version`


import numpy as np
np.__version__
#'1.16.4'

Numpy Basics

Numpy is a library that handles multidimensional array data. This is very useful when dealing with multiple arrays and matrices. If you understand the calculation of matrices, it will come in smoothly. (I was immature in the procession, so I was learning while looking at math textbooks. I sometimes calculate on paper.)

`Import Numpy`


import numpy as np

Numpy often translates to the module name ** np **. (I haven't seen any other module names.)

`Array definition`


a = np.array([1,2,3,4,5])
# []Don't forget

`output`


>>>a
array([1, 2, 3, 4, 5])
>>>print(a)
[1 2 3 4 5]
>>>type(a) #Type confirmation
numpy.ndarray

`When multiple arrays`


>>>b = np.array([[1,2,3,4,5],[6,7,8,9,10]])
>>>b
array([[ 1,  2,  3,  4,  5],
       [ 6,  7,  8,  9, 10]])
>>>print(b)
[[ 1  2  3  4  5]
 [ 6  7  8  9 10]]

`Display rows / columns`


>>>a.shape
(5,)
>>>b.shape
(2,5)

`np.Automatically create elements with arange`


>>>np.arange(1,10,2)
array([1, 3, 5, 7, 9])
>>>np.arange(2,11,2)
array([ 2,  4,  6,  8, 10])

`np.Convert the matrix with reshape`


>>>d = np.arange(1,11,1)
array([ 1,  2,  3,  4,  5,  6,  7,  8,  9, 10])
>>>d.shape #Display the shape of the array with shape
(10,)
>>>d.reshape(2,5) #2 rows x 5 columns matrix
array([[ 1,  2,  3,  4,  5],
       [ 6,  7,  8,  9, 10]])
>>>d.reshape(5,2) #5 rows x 2 columns matrix
array([[ 1,  2],
       [ 3,  4],
       [ 5,  6],
       [ 7,  8],
       [ 9, 10]])
>>>d.reshape(10,1) #10 rows x 1 column matrix
array([[ 1],
       [ 2],
       [ 3],
       [ 4],
       [ 5],
       [ 6],
       [ 7],
       [ 8],
       [ 9],
       [10]])

`np.You can also do this with arange and reshape`


>>>np.arange(1,10).reshape((3,3))
array([[1, 2, 3],
       [4, 5, 6],
       [7, 8, 9]])

Arithmetic calculation

You can also calculate between arrays. However, make sure that the array you are calculating matches the number of rows and columns.

`Arithmetic operations`


>>>x = np.array([1.0,2.0,3.0])
>>>y = np.array([4.0,5.0,6.0])
>>>x+y
array([5., 7., 9.])
>>>x*y
array([ 4., 10., 18.])
>>>x-y
array([-3., -3., -3.])
>>>x/y
array([0.25, 0.4 , 0.5 ])

broadcast

You can operate directly on the internal data of the array. The inner product always comes out in machine learning, so let's learn how it works. For inner products A and B, match the number of columns in A with the number of rows in B.

`Matrix multiplication(Inner product, dot product)`


>>>z = np.array([1,2,3])
>>>z.shape
(3,)
>>>v =np.array([[2],[3],[4]])
>>>v.shape
(3, 1)
>>>np.dot(z,v)
array([20])

`inner product`


>>>e = np.array([[1,2,3],[4,5,6]])
>>>e
array([[1, 2, 3],
       [4, 5, 6]])
>>>f =np.array([[3,3],[3,3],[3,3]])
>>>f
array([[3, 3],
       [3, 3],
       [3, 3]])
np.dot(e,f)
array([[18, 18],
       [45, 45]])

Finally

This time I wrote a simple part of Numpy. Next, I will write the mechanism of neural network calculation using matrices.

References

Deep Learning from scratch
Learn by running with Python! A new machine learning textbook
New data analysis textbook with Python

[PYTHON] Why what? Deep Learning Scientific Calculation Library Numpy Edition