A Brief Introduction to Machine Learning-2: Learning and Problem Types in Machine Learning

Melis Tekin Akcin
5 min readSep 18, 2023

--

“Machine learning is the science of getting computers to learn and act like humans do, and improve their learning over time without being explicitly programmed. — Arthur Samuel”

The primary aim of this article is to ensure that readers are on the same page with the fundamental concepts of machine learning.

I want to explain these concepts in a simple and clear way so that whether you’re new to the topic or have some knowledge already, you can easily follow along.

My focus is to create a shared understanding of what machine learning is, how it works, and where it’s used. This foundation will help you get started or refresh your knowledge in the fields of artificial intelligence and data science and serve as a basis for upcoming articles in our machine learning series.

Furthermore, I firmly believe that having a solid understanding of variable types, learning types, and problem types is crucial in simplifying the machine learning process. Understanding these fundamental aspects ensures that you’re starting off on the right foot with your ML project.

Firstly, let’s answer the following question:

What is the difference between classical programming and machine learning?

Programming

Programming involves writing a set of explicit instructions or algorithms that tell a computer how to perform a specific task or series of tasks.

You are giving the input and the algorithm to your computer, you are the one who determine the rules and the logic and the computer produces the output.

Machine Learning

Machine learning focuses on training algorithms to learn patterns and make predictions or decisions based on data.

You are giving the input which is your data and the output, and the computer discovers the patterns in your data and learning from the examples. The computer is the one who produces the algorithm based on the data.

The more opportunities you give a computer to learn, the better your algorithm gets. In machine learning, the computer learns by the data and you don’t need to program explicitly.

The following visual makes the difference pretty clear!

Learning Types

Photo Credit: https://analyticslearn.com/

Machine learning includes different ways of learning that help computers tackle a wide range of tasks by looking at data.

1. Supervised Learning:

In supervised learning, you have input (independent variables, denoted as X) and a target (dependent variable or label, denoted as y).

Each input is associated with a corresponding output. In other words, you are also giving the algorithm the “right” answers.

Problem Types for Supervised Learning:

Regression

Regression problems involve predicting a numerical target.

For example, predicting the price of a house based on its features.

Classification

Classification problems involve assigning data points to predefined categories or classes.

For example, predicting whether an email is spam or not is a type of classification problem. For instance, identifying malignant and benign tumours in breast cancer detection.

2. Unsupervised Learning:

Photo credit: cronj.com

In unsupervised learning we are given data, but this case we do not have the “right” answers. In other words, we have the input features, but these are not associated with an output (y).

Unsupervised learning involves learning patterns and relationships in data without labeled outputs.

Problem Types for Unsupervised Learning

Clustering:

Clustering problems involve grouping the data points based on their characteristic and the patterns on the data set.

For example, grouping the customers based on their purchase behaviours, grouping similar movies for recommendation, or content categorization for websites like Google News.

Dimensionality Reduction:

The aim for dimension reduction is to reduce the number of features in your data while also representing it effectively. It can be achieved by using Principal Component Analysis (PCA) which we will cover in the upcoming articles.

Anomaly Detection:

Anomaly detection problems focus on identifying unusual data points, such as credit card fraud detection.

3. Semi-Supervised Learning:

Semi supervised learning is, as the name suggests, the combination of supervised and unsupervised learning.

In real life data, the data is not always as we wanted. For example, we may have a target variable but with a few labeled but a larger amout unlabeled data. In this cases, we may use semi-supervised learning.

4. Reinforcement Learning:

In reinforcement learning, agents learn to make a sequence of decisions by interacting with an environment to maximize a cumulative reward.

The agent learns from its mistakes, receiving rewards for correct choices and punishments for bad ones. For example, training a computer to play chess.

I hope you enjoy reading!

In the upcoming articles, we will delve deeper into these concepts and explore more advanced topics in the fascinating world of machine learning!

Stay tuned! :)

Note: If you enjoy reading so far, maybe you may enjoy with my other contents! You may follow for more articles.

Happy analysing,

Melis.

References

  1. https://www.miuul.com/
  2. Coursera & DeepLearningAI, Supervised Machine Learning: Regression and Classification
  3. https://medium.com/@meliss85/a-brief-introduction-to-machine-learning-1-5407d47558d1?sk=1b8b7f8d5f4244fccda1388365bae58b

--

--

Melis Tekin Akcin

PhD on Mathematics & Data Analyst & Passionate about learning on Data Science, Machine Learning, Artificial Intelligence and Math & Love writing