All the code examples in this post are written and executed in a Jupyter Notebook.

NumPy, which is short for Numerical Python, is a Python library optimized for numerical computing.

At the core of NumPy is a high-performance data structure for representing multidimensional arrays known as ndarray. NumPy arrays are homogeneous, meaning all elements have the same data type.

Creating NumPy arrays

First, we import the numpy module, which by convention, is imported under the alias np.

Import numpy as np

We can create arrays (ndarray instances) using several functions provided by the NumPy library.

Creating arrays from lists

We can create a one-dimensional array by passing a Python list as an argument to the np.array function.

data = np.array([1, 2, 3, 4])

data
# array([1, 2, 3, 4])

type(data)
# numpy.ndarray

To create a two-dimensional array, we simply pass a nested Python list to the np.array() function.

data = np.array([[1, 2], [3, 4]])

data
# array([[1, 2],
#        [3, 4]])

Creating arrays from constant values

np.zeros

We can create an array filled with zeros using the np.zeros function. The first argument that we pass to np.zeros() can be an integer or a tuple that describes the number of elements along each dimension of the array.

data = np.zeros((2, 2))

data
# array([[0., 0.],
#        [0., 0.]])

np.ones

We can create an array filled with ones using the np.ones function. The first argument that we pass to np.ones() can be an integer or a tuple that describes the number of elements along each dimension of the array.

data = np.ones((4))

data
# array([1., 1., 1., 1.])

By default, np.zeros() and np.ones() create arrays with elements of data type float64. We can convert the default data type to any other data type of our choice by passing an optional keyword (dtype) argument.

data = np.ones(4, dtype=np.int64)

data
# array([1, 1, 1, 1], dtype=int64)

data.dtype
# dtype('int64')

We can create an array filled with any arbitrary constant value by first creating an array filled with ones and then multiplying the array with the fill value.

data = np.ones((2,3)) * 5

data
# array([[5., 5., 5.],
#        [5., 5., 5.]])

np.full()

We can produce the array above using the np.full() function as well.

data = np.full((2, 3), 5)

data
# array([[5, 5, 5],
#        [5, 5, 5]])

Creating arrays filled with incremental sequences

We can create an array with evenly spaced values between a start value and an end value using the np.arange() function. The third argument is the number we want the values to increment by.

data = np.arange(1, 10, 3)

data
# array([1, 4, 7])

Creating an array with random numbers

We can create an array with random numbers that are uniformly distributed between 0 & 1 using the np.random.rand() function.

data = np.random.rand(5)

data
# array([0.32108486, 0.7915518 , 0.44758901, 0.95538267, 0.56497222])

Basic attributes of the ndarray class

The ndarray class provides several attributes for finding out important metadata about an array.

• shape: A tuple that contains the number of elements for each dimension of the array.
• size: The total number of elements of the array.
• ndim: Number of axes in the array
• nbytes: No. of bytes used to store the data
• dtype: The data type of elements in the array.

In the code example below, we have created a NumPy array using the np.array function on a python list.

data = np.array([[12, 24], [15, 25], [20, 40]])

data.shape
# (3, 2)

data.size
# 6

data.ndim
# 2

data.nbytes
# 24

data.dtype
# dtype('int32)