Machine Learning Diploma
Session 2: Numpy
Agenda:
1 Numpy Basics
2 Quick Array Creation
3 NPZ Files
4 Stacking
5 Comparison
6 Conditional Access & Modification
7 Array BroadCasting
8 Sparse Matrices
1. Numpy Basics
What is Numpy?
➢ One of the main primarily used data structures to
represent & process data.
➢ Numpy is about creating N-dimensional Tensors, these
Tensors are called Numpy Arrays or ndarray.
➢ Numpy arrays could be 0-dimensional(scalars), 1-
dimensional(vectors), 2-dimensional(matrices), 3-
dimensional(Images), etc.
Import:
Create Numpy Arrays:
Dimensions & shape:
Reshape:
Add Dimension:
Dtype:
➢ Numpy array can only carry one data type.
➢ If you pass elements with different datatypes, Numpy will
change all of them into one datatype (the most general
dtype).
➢ Here is the order of general datatypes:
➢ Object > String > Float > Int > Bool.
Dtype examples:
Change Dtypes:
➢ Numpy allow you to Change the Datatype of ndarrays.
➢ Examples:
Indexing:
➢ Means accessing one element in Numpy array using its index.
Slicing:
➢ Means accessing many elements in an array using index range.
Transopose:
➢ Make matrix columns become rows and rows become
columns.
2. Quick Array Creation
Quick arrays Methods :
➢ Numpy provides you some built-in methods that helps you
create arrays quickly.
➢ In the coming slides, you will find the most popular
methods.
Zeros():
➢ Creates an array of the specified size with the contents
filled with zero values.
Ones():
➢ Creates an array of the specified size with the contents
filled with one values.
Full():
➢ Creates a new array of the specified shape with the
contents filled with a specified value.
Empty():
➢ Allocates space of a numpy array in the memory without
initializing array elements values.
➢ Use it when you want to create arrays quickly if you plan
to fill them with meaningful values later.
arange():
➢ Generates values starting from a given start value,
incrementing by a step.
➢ It takes three parameters; (start, stop, step).
Linspace():
➢ Is used to create an array of evenly spaced values within a
specified range, For example, np.linspace(0, 10, num=20) will
create an array of 20 evenly spaced values between 0 and 10,
including both 0 and 10.
Linspace():
➢ To create 10X10 Matrix using Linspace:
Identity ():
➢ Creates an identity matrix with a given shape.
Random.random():
➢ Creates an array of a given shape with random values
between 0 & 1.
Random.randint():
➢ Creates an array of a given shape with random values
between a & b, where a & b are boundaries that you
specify.
Random.choice():
➢ Creates an array of a given shape with random values
sampled from elements of another array.
3. NPZ files
What are NPZ files?
➢ Numpy provides a way to save your numpy arrays as files
called npz files which helps you to save the data you want
in npz format.
➢ You can save the files using a built-in method called savez,
and you can load using a built-in method called load.
Save Numpy arrays: Load Numpy arrays:
4. Stacking
What Is Stacking?
➢ Stacking means Concatenating two matrices together.
➢ There are two types of Stacking:
Vertical Stacking Horizontal Stacking
➢ The two matrices must have ➢ The two matrices must have
the same number of columns the same number of rows.
Why Stacking?
➢ Sometimes you collect datasets from different sources, and
you might want to concatenate them into one dataset.
Vertical Stacking Horizontal Stacking
5. Comparison
Comparison using ‘==’:
➢ Is about elementwise comparison between two numpy arrays.
➢ The result is a Boolean numpy array with the same shape.
➢ Examples:
Comparison using all():
➢ Returns True if the all elements follow the condition
Comparison using any():
➢ Returns True if any element follows the condition
Comparison using isclose():
➢ It applies elementwise comparison with tolerance, where
we check if every element is equal or close to its
corresponding element in another array.
➢ In the following example, tolerance = 1.
Comparison using allclose():
➢ It applies comparison with tolerance, where we check if all
elements in an array is equal or close to its corresponding
element in another array.
➢ In the following example, tolerance = 1.
6. Conditional Access & Modification
Conditional Access:
➢ It means getting elements of an array that satisfy a specific
condition.
➢ A Condition means the result of elementwise comparison, we
saw in the previous section.
Condition Conditional Access
Conditional Modification using:
➢ It means applying modification to elements of an array that
satisfy a specific condition.
Condition Conditional Modification Conditional Modification
using ‘==’ using where()
7. Array BroadCasting
Array BroadCasting:
➢ Is the application of arithmetic operations between arrays
with a different shape.
Vector/Scalar BroadCasting Matrix/Vector BroadCasting
8. Sparse Matrices
Sparse Matrix:
➢ Sparse matrix is a matrix that contain mostly zero values.
➢ The opposite of sparse matrix is dense matrix, which is a
matrix where most of the values are non-zero.
➢ Sparse matrices has a problem related to waste of memory
resources as those zero values do not contain any
information.
Sparse Matrix:
➢ The solution is to transform it into another data structure.
Where the zero values can be ignored.
➢ There are two techniques:
➢ These two techniques are doing the same thing, but in
different ways, which we are not interested in.
Thank You
