See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/339031674

Getting Started with Machine Learning (ML)

Article · February 2020

CITATION READS

1 3,747

1 author:

Rukshan Manorathna
University of Colombo
16 PUBLICATIONS 9 CITATIONS

SEE PROFILE

All content following this page was uploaded by Rukshan Manorathna on 04 February 2020.

The user has requested enhancement of the downloaded file.

Getting Started with Machine Learning (ML)
A disruptive technology which sweeps away traditional programming in certain cases…

Rukshan Pramoditha
Dec 25, 2019 · 9 min read

After a very long time since the first article Getting Started with Data Science was
published in Data Science 365, today, I meet you with a new topic in the field of Data
Science. If you haven’t read my first article yet, I recommend you to read it before reading
this one. Without further delay, welcome to Getting Started with Machine Learning (ML)!

Introduction
Machine Learning (ML) is a collection of algorithms & techniques used to build
systems that learn from data. These systems are then able to perform predictions by
finding patterns in data. Machine learning is a disruptive technology which sweeps away
traditional programming in certain cases. What is the difference between machine learning
& traditional programming?
In traditional programming, the data & the program produce the output. For example, when
performing some accounting tasks, the program takes in the data (sales records, inventory
lists, etc.) & calculates your profits or losses. It will also give some nice & fanciful charts
showing your sales performance.

Traditional programming is a manual process — meaning a person (programmer) creates the

program by manually formulating or coding the rules.

In machine learning, the data & the output produce the program. You take both data &
output & use them to derive a set of rules to make predictions. It means that you may use
this model to predict the most popular items that will sell next year.

In machine learning, the algorithm automatically formulates the rules from the data.

ML consists of the following disciplines:

• Scientific computing (e.g. Python or R programming)

• Mathematics
• Statistics

When using ML, a data scientist needs to know about the followings:

• Which method of machine learning will best help in completing the given task
• How to apply that method

The knowledge that how that method works is optional.

Why machine learning?
Consider an event of creating a spam e-mail filter program without using ML. The
programmer writes that program by following the steps in a certain order.

Step 1: Identify how spam e-mails look like. (Words: “debit card”, “free”, “for you”, etc.)

Step 2: Write an algorithm to detect the patterns that you’ve seen.

Step 3: Finally, you’d test the program & then make changes to the program until the results
are good enough.

After that, the software would flag emails as spam if a certain number of those patterns are
detected.

In your algorithm, you write rules in your program to detect the patterns. If this is a case
that requires a long list of rules to find the solution, it soon becomes difficult for a human to
accurately code the rules. You can use ML to effectively solve this problem.

What will happen if the e-mail senders change their e-mail templates so that a word like
“4U” now instead of “for you”? The program using traditional techniques would need to be
updated manually. On the other hand, a program using ML techniques will automatically
detect this change & will adapt to new data.
When should you use machine learning?
ML is not a solution for every type of problem. There are certain cases where solutions can
be developed without using ML techniques. For example, you don’t need ML if you can
determine a target value by using simple rules & computations. If you implement ML
techniques for such problems, they will be getting more complex & you will need more
computer power to get the solutions.

However, there are some situations where you need to use ML.

• Very complex problems for which there is no solution with a traditional

approach: You need to use a data-driven approach to get the solution. For example,
speech recognition: When you say “one” or “two”, the program should distinguish
the difference. You will need to develop an algorithm that measures sound.

• Non-stable environments: Machine learning software can adapt to new data.

• When you have a problem that requires many long lists of rules to find the
solution: When rules depend on too many factors, it soon becomes difficult for a
human to accurately code the rules. You can use ML to effectively solve this
problem.
• You cannot scale: ML solutions are effective at handling large-scale problems. For
example, a spam e-mail filter program: There are millions of emails which are needed
to check in a single day! It cannot be done manually.

A few of ML terminology
There is a long list of ML terminology. Here I list some terms which are mostly used. Others
will be given in relevant parts as we progress.

• In ML, a target is called a label. In statistics, a target is called a dependent

variable.
• A variable in statistics is called a feature in ML.
• A transformation in statistics is called a feature creation in ML.

Types of machine learning

There are 2 main types of Machine Learning which are based on their learning styles. They
are:

• Supervised Learning
• Unsupervised Learning

In addition to these types, there are other types as well:

• Semi-supervised Learning
• Reinforcement Learning
• Batch Learning
• Online Learning
• Instance-based Learning
• Model-based Learning
The criteria for the above classification are:

• According to the type & amount of human supervision needed during the
training: Supervised Learning, Unsupervised Learning, Semi-supervised Learning &
Reinforcement Learning are classified based on this criterion.
• If they can learn incrementally
• If they work simply by comparing new data points to find data points or can
detect new patterns in the data & then will build a model

Supervised vs unsupervised learning

Supervised Learning
In supervised learning, you supervise the learning process. You train the algorithm using
labelled data. You can correct your algorithm if it makes a mistake in giving you the answer
in the learning process. This can be compared to learning which takes place in the presence
of a supervisor or a teacher.

In supervised learning, both input & output variables are given. This can be explained
mathematically. You have input variables (called x) & an output variable (called y) & you
use an algorithm to learn the mapping function of from the input to the output, y=f(x). In
supervised learning, the goal is to determine the mapping function so well that when
you have new input data (x), you can predict the output for that data. If the mapping is
correct, the algorithm has successfully learned. Else, you make the necessary changes to the
algorithm so that it can learn correctly. In simple words, what we do in supervise learning is
that we use a labelled dataset to obtain a new label for unlabelled data.
There are two phases: Training Phase & Testing Phase. In the training phase, you take a
randomly selected specimen of geometric shapes (training data) & label them accurately.
Then you make a table of all the characteristics (features) of each shape. You feed this data
to the machine learning algorithm & it learns a model (called prediction model). In the
testing phase, you input a shape (test data) which was also labelled. The prediction model
which was created earlier will give the correct label for that shape. If the output is correct,
the algorithm has successfully learned. Else, you make the necessary changes to the
algorithm so that it can learn correctly.

Classification and Regression are two types of supervised learning. Supervised learning is
a highly accurate method. The main drawback is that classifying big data can be a real
challenge.

Unsupervised Learning
In unsupervised learning, you do not need to supervise the learning process. Instead, you
need to allow the model to work on its own to discover hidden patterns in data. This can be
compared to the learning process of a student who has textbooks and all the required
material to study but has no teacher to guide so that he will have to learn by himself.
In unsupervised learning, only input data will be given. That data does not have any sort of
labels. The goal is to find the hidden patterns or the underlying structure in the given
input data in order to learn about the data. Unsupervised learning helps you to find
features which can be useful for categorization. It can also help to detect anomalies and
defects in the data which can be taken care of by us.

Clustering and Association are two types of unsupervised learning. Unsupervised learning
is less accurate & computationally complex when compared to supervised learning.

As a summary, we can compare supervised & unsupervised learning methods as follows.

Machine learning algorithms

An algorithm is a set of rules that a machine follows to achieve a particular goal. These
rules are in a certain order. A learner or machine learning algorithm is a set of rules used
to learn a machine learning model from data. ML algorithms are given general guidelines
that define the model, along with data while classical algorithms are given exact & complete
rules to finish a task. An ML algorithm can accomplish its task when the model has been
adjusted with respect to the data. We have to fit the model on the data or the model has to
be trained on the data. ML algorithms are different from classical algorithms in a way that
they automatically learn from the data you provide.

Some useful definitions of commonly used terms related to machine learning algorithms are:

Dataset: A table with the data from which the machine learns. The dataset contains the
features and the target to predict.

Instance: A row in the dataset. Other names for “instance” are data point, observation.
An instance consists of the feature values x(i) & if known, the target outcome y(i).

Feature: An input used for the machine learning algorithm. A feature is a column in the
dataset. The matrix with all the features is called X for a single instance. The vector of a
single feature for all instances is x(j).

Target: The information that the machine learns to predict. In mathematical formulas,
the target is usually called y. In statistics, it is called a dependent variable.

Prediction: The target value that the machine learning model “guesses” based on the
given features.

Machine learning algorithms fall into two broad categories which are supervised learning
algorithms & unsupervised learning algorithms, although there are other categories such
as semi-supervised learning algorithms, reinforcement learning algorithms, etc. The
following chart shows some of the commonly used algorithms which are classified under
supervised & unsupervised. Note that some of them can be in both supervised &
unsupervised categories, although they are listed under one category.
What is next?
This is just an introductory article for ML. This article lays a good foundation for ML &
motivates you to learn more about ML. The next big part is to learn machine learning
algorithms. Learning theoretical parts of these algorithms is not just enough. We also want
to learn how to implement these algorithms using Python or R programming. Here I use
Python. (Selecting Python or R for doing Data Science & ML should be done by conducting
thorough research. I personally selected Python by comparing many factors between the
two languages.)

The next article will be logistic regression under classification algorithms in ML. An article
series about Python programming should also be written parallelly with the ML article
series.
One last key point
My success will not be possible without your feedback. So please don’t hesitate to give me
feedback. Write them in the comment section of this article or just drop a message at
[email protected].

Thank you for reading! Next time, I will meet you with another ML article. Goodbye for
now!

Written by: Rukshan Pramoditha

Data Science 365,

Bring data into actionable insights.

View publication stats