Numpy - Intro

  • Backbone for all scientific computations in Python.
  • Multidimensional array.
  • Provides lot of predefined routines for matrix operations

1. How to install numpy and import?

from shell/command prompt run one of the following commands

1.1 npm

npm install numpy

or

1.2 conda

conda install numpy

In [1]:
import numpy as np
# np is the alias for numpy

2. Creating a numpy array

2.1 Two dimensional array

In [2]:
# Create a regular two dimensional array

arr_2d = [
    [4, 2, 6],
    [5, 1, 3],
    [9, 8, 7]
]

np_arr_2d = np.array(arr_2d)

print(np_arr_2d)

type(np_arr_2d)

[[4 2 6]
 [5 1 3]
 [9 8 7]]
Out[2]:
numpy.ndarray
In [3]:
arr_1d = [4, False, 5, 9, True] # Mixed data types

np_arr_1d = np.array(arr_1d)

print(np_arr_1d) # All converted to same data type - True = 1, False = 0

[4 0 5 9 1]
In [4]:
arr_1d = [4, 'Hello', 5, 9, 'World'] # Mixed data types

np_arr_1d = np.array(arr_1d)

print(np_arr_1d) # All converted to string with quotes

['4' 'Hello' '5' '9' 'World']

2.2 Frequently used methods in numpy

In [5]:
# Let's create an array

arr_3d = [                         # 4 x 3 x 3
    [                              # Row 1
        [ 3, 24,  1],              #
        [69,  8, 10],              # 3 x 3 array
        [57, 21,  9]               #
    ],[                            # Row 2
        [ 1, 19, 62],              #
        [25, 23, 28],              #
        [27, 96,  1]               #
    ],[                            # Row 3
        [65, 24, 11],              #
        [32,  1, 84],              #
        [ 2, 43,  1]               #
    ],[                            # Row 4
        [14, 81, 12],              #
        [19,  9, 21],              #
        [ 18, 8, 21]               #
    ]
]

np_arr = np.array(arr_3d)
In [6]:
np_arr.shape # Shape of the array 3x3x3
Out[6]:
(4, 3, 3)
In [7]:
np_arr.dtype # data type of elements in the array
Out[7]:
dtype('int32')
In [8]:
np_arr.size # total number of elements in the array = 4 * 3 * 3 = 36
Out[8]:
36
In [9]:
np_arr.ndim # number of dimensions in the array
Out[9]:
3
In [16]:
# Generating a zero array

z_arr1 = np.zeros(3)
z_arr2x2 = np.zeros( (2, 2) ) # [2x2] dimension

print(z_arr1)
print('--------------- 2 x 2')
print(z_arr2x2)

[0. 0. 0.]
--------------- 2 x 2
[[0. 0.]
 [0. 0.]]
In [11]:
# Generate an array of ones
# Read the arrays bottom up (3x4) x 2
np.ones( (2,3,4) )
Out[11]:
array([[[1., 1., 1., 1.],
        [1., 1., 1., 1.],
        [1., 1., 1., 1.]],

       [[1., 1., 1., 1.],
        [1., 1., 1., 1.],
        [1., 1., 1., 1.]]])
In [12]:
# Generate a number sequence in array with an interval
np.arange( 10, 30, 2 )
# from 10 to < 30 with step size 2
Out[12]:
array([10, 12, 14, 16, 18, 20, 22, 24, 26, 28])
In [13]:
# Creating n number of divisions between two given numbers - inclusive of the first and last number
np.linspace( 0, 12, 5 )
# Create 5 equidistant numbers from 0 to 12 inclusive
Out[13]:
array([ 0.,  3.,  6.,  9., 12.])
In [17]:
# mathematical operations
np.round( np.sin ((22/7) / 2) ) # sin ( pi/2 )
Out[17]:
1.0
In [26]:
# Saving an array to a text file

arr_3x3 = [
    [ 1,  2,  5 ],
    [ 4,  3,  5 ],
    [10, 14, 18]
]

np_3x3 = np.array(arr_3x3)

np.savetxt('test.csv', np_3x3)