Numpy Module - Arrays, 2D Arrays

numpy Module
numpy is a Python module short for Numerical Python used for scientific computing
and data analysis.
numpy provides:
• Multi-dimensional arrays (data structures) for fast, efficient data storage and
manipulation.
• Useful functions to manipulate arrays of data.
• Tools for importing and exporting data to and from programs.

2
 numpy Package
numpy functionality used for data analysis includes:
• Array operations for data manipulation.
• Common algorithms such as sorting, unique and set operations.
• Descriptive statistics, aggregating and summarizing data.
• Merging and joining data sets.
• Expressing conditional logic as array expressions instead of loops.

3
Using numpy Functionality
To use the numpy package, it must first be imported as before.
◦ import numpy as np -> when we want to use the functionality in the package, we
will do so using the syntax: np.functionName(…)

4
numpy Data Structure: ndarray
Stores multiple values together in a data structure.
It is a fixed-sized array in memory that contains data of the same type.
Unlike lists, all elements must be of the same type, an array cannot contain mixed types.
The size of a numpy is fixed when it is created, an array cannot be made larger or smaller;
appending/inserting/deleting elements is not permitted.
Numpy arrays are mutable, the values stored in the array can be changed.
Have built in functionality to allow element by element operations, without using loops.

5
numpy Array Attributes
A Numpy array is a row/grid of values, all of the same type, and is indexed by a tuple of
non-negative integers.
Each array has the following attributes:
◦ dtype - the type of data in the array
◦ ndim - the number of dimensions,
◦ shape - the size of each dimension, and
◦ size - the total number of elements in the array

6
Functions to Create ndarrays
There are several functions that are used to create numpy arrays, these include:
• array(): creates an array with the given sequence of values.
• arange(): creates an array with a specified range of values.
• linspace(): creates an array with n linearly spaced values.
• empty(): creates an array of a given shape with empty elememts.
• zeros(): creates an array with a given shape filled with zeros.
• ones(): creates an array with a given shape filled with ones.
• rand() /randint(): creates an array with a given shape filled with randomly generated values.

7
Creating numpy Arrays
We can initialize numpy arrays from Python lists, and access elements using square brackets.
To create an array from a list, we can use the array() command, and pass the list as a parameter.
Example – create a one-dimensional array:
import numpy as np
z = np.array([4,7,6,9,2])

z
Out[13]: array([4, 7, 6, 9, 2])

print(z)
[4 7 6 9 2]

8
Examples with Attributes
r = np.array([87,33,5,16,9,17])

r
Out[29]: array([87, 33, 5, 16, 9, 17])

r.dtype
Out[30]: dtype('int32')

r.ndim
Out[31]: 1

r.shape
Out[32]: (6,)

r.size
Out[33]: 6

9
 Creating numpy Arrays - arange
Numpy arrays can also be generated using the arange function.
The arange function takes the starting, ending and step values and generates an array containing
the given values.
Example:
b = np.arange(1,10,2)
b
Out[67]: array([1, 3, 5, 7, 9])
t = np.arange(1,10)
t
Out[65]: array([1, 2, 3, 4, 5, 6, 7, 8, 9])

10
linspace()
Function creates evenly spaced float values.
Takes start, end and number of values as parameters.
Generates the given number of values between the specified interval.

11
Creating Special Arrays
We have already used the array and arange functions to create arrays with initial values.
There are also special functions we can use to create arrays with other default values (ones, zeros).

Function Output
array Creates a numpy array object
arange Creates a numpy array with values within the specified range (start,stop,step)
ones Creates a numpy array containing all ones, with the given shape.
zeros Creates a numpy array containing all zeros, with the given shape.

12
Creating 1-D arrays: zeros(), ones()

13
 Generating Random Arrays
In numpy there is a sub-module, random, that contains functions for generating random arrays.
For example:
◦ rand( n ) – generates an array of n random values between 0-1.
◦ randint( ) – generates an array of n random integer values between low (inclusive) and high (exclusive).

import numpy as np
r_vals = np.random.rand(10)
print(r_vals)
r_vals = np.random.randint(1,26,12)
print(r_vals)

14
 Accessing and Updating numpy Arrays
Accessing elements:
z[0] -> 4
z[3] -> 9
Updating elements:
z[2] = 5
z
Out[19]: array([4, 7, 5, 9, 2])

15
Slicing numpy Arrays
Like with lists, we can access parts of a numpy array using the start:stop:step syntax.
Reminder – the stop value is up to, but not inclusive of the element specified.
If any of the values are omitted, the default values will be start: 0, stop: end, step: 1.
Examples:
z
Out[22]: array([4, 7, 5, 9, 2])
z[1:4:2]
Out[21]: array([7, 9])
z[1:4]
Out[23]: array([7, 5, 9])
z[1::2]
Out[24]: array([7, 9])
z[:4]
Out[25]: array([4, 7, 5, 9])

16
Creating Two-Dimensional Arrays
Two dimensional numpy arrays are just arrays of arrays.
We created the array using the array command and passing a list as a parameter, now we can
pass a list of lists.
Example:
m = np.array([[3,5,2,4],[7,6,8,8],[1,6,7,7]])
m
Out[35]:
array([ [3, 5, 2, 4],
[7, 6, 8, 8],
[1, 6, 7, 7]])

17
Creating 2-D arrays: zeros(), ones()

18
Accessing Elements: Two D Arrays
Syntax: arrName[ row, col]

Examples:
#accessing elements
m[1,2]
Out[43]: 9
#updating elements
m[2,1] = 12
m
Out[45]:
array([[ 3, 5, 2, 4],
[ 7, 6, 9, 8],
[ 1, 12, 0, 7]])

19
Accessing Elements - Slicing
We can also slice 2D arrays, using the start:stop:step syntax for each dimension (row, col) separated by
a comma.
Examples:
m[ 1:2,1:2]
Out[46]: array([[6]])
m[-1,-1]
Out[48]: 7
y = m[2, :]
y
Out[50]: array([ 1, 12, 0, 7])
t = m[2, ]
t
Out[52]: array([ 1, 12, 0, 7])
w = m[2:0:-2,1:3]
w
Out[56]: array([[12, 0]])

20
Element-Wise Operations
Unlike lists, arrays allow us to perform element by element operations without using a loop.
When we apply arithmetic operations to arrays, the operation will be applied to each element
of the array.
See: 10_OneDelementOperations.py
See: 10_TwoDelementOperations.py

21
Python Arrays – Mathematical
Functions
Basic mathematical functions operate element wise on arrays and are available both as
operator overloads and as functions in the numpy module.

Basic mathematical functions operate element wise on arrays and are available both as
operator overloads and as functions in the numpy module.
See: 10_numpyArrayMath.py

22
 Numpy Arrays – built in functions
Numpy arrays have many built in functions for computing statistics on an array.
See: 10_numpyArrayFunctions.py

Function Name Description

np.sum() Computes the sum of values in given array.
np.mean() Computes the mean(average) value in given array.
np.max() Computes the maximum value in given array.
np.min() Computes the minimum value in given array.
np.argmax() Computes the index of the maximum value in given array.
np.argmin() Computes the index of the minimum value in given array.

23
Numpy Arrays – Reshaping and Transpose
Another useful type of operation is reshaping arrays. This can be done using the reshape() function.
Example: reshape x = np.arange(1,10) into a 3 x 3 array:

You can also do common matrix operations such as finding the transpose. The transpose of a matrix switches the row
and column indices to produce another matrix.
Example: compute the transpose using .T

24
numpy Arrays – Matrix Operations
numpy knows how to efficiently do typical matrix operations
Examples:
a matrix-vector product using np.dot:

25
numpy Arrays – Concatenation
It’s also possible to combine multiple arrays into one, and to conversely split a single array into
multiple arrays.
Concatenation, or joining of two arrays in NumPy, is primarily accomplished through the routines
np.concatenate, np.vstack, and np.hstack.
np.concatenate() - takes a tuple or list of arrays as its first argument, returns a new array.

Output:

26
numpy Arrays – Concatenation
np.concatenate() – can be used to concatenate more than two arrays or two-dimensional
arrays as well.

Output:

27
numpy – hstack, vstack
The concatenate function can be used when the arrays you are joining have the same
number of dimensions (one or two).
If you want to join arrays with different dimensions, you should use the functions vstack or
hstack.
vstack: stack two or more arrays vertically and returns a new array
hstack: stack two or more arrays horizontally and returns a new array.
See: 10_concatenate_vstack_hstack.py

28
numpy Arrays – Splitting
The opposite of concatenation is splitting, which is implemented by the functions np.split,
np.hsplit, and np.vsplit.
For each of these, we can pass a list of indices giving the split points.

29
 Sorting Arrays
Like lists, numpy arrays can be sorted in place using the sort() function.

30
Unique and Set Logic
For one dimensional arrays, numpy has some basic set operations.
The most used is the unique()function, which returns the sorted unique values in an array.

31
 Boolean Arrays
A Boolean array is an array that contains True/False values.
A Boolean array may be generated in numpy by using element-wise relational expressions
with arrays.

Python assumes that True values are 1 and False values are 0. If we sum() the elements of
a Boolean array, the True values will be counted.

32
 Boolean indexing and sub arrays
We can use Boolean arrays to select a subset of an array.
For example if we want to find the rainfall for all days of the week in which it was greater than
10, we can do so using logical indexing.

33
where() function – array indexes
In the previous example we used the logical array to select the values/elements that met the given condition(s).
If we want to find the indexes of the elements that meet the given condition(s), we can do so using the where
function.
The example below shows that the resulting values are the indexes of the elements in the array that meet the
given condition. I.e. elements 0,1,2, etc have values over 10.

34
Subarrays and functions
import numpy as np
x = np.array([22,41,58,33,17,32])
#returns array of bool values for each element in x where condition is true
r1 = x > 40
print(r1)

#returns array containing elements in x where condition is true

r2 = x[x > 40]
print(r2)

#returns array containing indices of elements in x where condition is true

r3 = np.where(x > 40)
print(r3)

35
Reading data from files with numpy
numpy contains functions that allow us to read text files into numpy arrays.
loadtxt(filename)
Reads data from a file (in the current directory unless otherwise specified) into a numpy
array.
Default delimiter is whitespace.
Default type is float.
We can add additional parameters when reading the file. For example, if the first row in the file contains
headers, we would use the skiprows parameter to start at the second row.
loadtxt(‘myfile.txt’,skiprows=1)
To read string data from the file use the dtype parameter:
loadtxt(‘myfile.txt’,dtype=‘str’)
See loadtxt documentation for a full list of parameters.
36
Writing Data to Files with numpy
Numpy contains functions that allow us to write numpy arrays to text files.
savetxt(filename, arrayName)
◦ Writes the data in the numpy array to a text file.
◦ By default, values in each row are space delimited, and rows are delimited by newlines.
We can add additional parameters when writing data to a file.
See savetxt documentation for a full list of parameters.

37
Example
The file 10_world_pop_area.txt contains the population and area for a list of countries. The
first row in the file contains headings.
The file 10_countries.txt contains the text country names.
Write a program that does the following:
◦ reads the population and areas from the file into an array.
◦ Read the country names into a separate array.
◦ Increases the population (in place) by 10%
◦ Calculates the population density for each country (population per km2) and stores in a new array.
◦ Writes the countries and densities to a new file, densities.txt.
◦ Finds and displays the name of the country with the maximum density, together with the maximum density.

See: 10_numpyFiles.py

38
Terms of Use
⮚ This presentation was adapted from lecture materials
provided in MIT Introduction to Computer Science and
Programming in Python.
⮚ Licenced under terms of Creative Commons License.

39