Demystifying Machine Learning

Intro

In today’s tech-driven world, machine learning (ML) has become a buzzword that seems to pop up everywhere, from online shopping recommendations to voice-activated assistants, and of course for doing your homework or TOK assignments (just kidding). But what exactly is machine learning, and how does it work? In this beginner’s guide, I’ll break down the basics of machine learning and hopefully make you understand it in a very simple way!

What is Machine Learning?

Let me geek out a little bit first, machine learning is a subset of artificial intelligence (AI) that empowers systems to learn from data, identify patterns, and make decisions with minimal human intervention. Unlike traditional programming—where developers painstakingly write out explicit instructions for every task (and get a small syntax error that makes them question everything)—machine learning flips the script. Here, algorithms learn from experience! This is where the magic happens, often referred to as unsupervised learning (cool, right? Don’t lose me yet!).

This incredible ability to learn from data makes machine learning especially powerful when tackling complex problems. Imagine teaching a computer to recognize faces in photos or predicting what movie you’d like to watch next, all without telling it how to do it step by step. That’s the beauty of machine learning!

So Machine Learning isn't really that of new of technology, It was actually created in 1950 by the one and only GOAT of computer science Alan Turning, he published it in his article “Computer Machinery and Intelligence” which eventually became The Turing Test, which experts used to measure computer intelligence. Wait but why didnt we hear until a couple of years back since the surge of Large Language Models like chatGPT (General Pre-trained transformer)?

Well, unfortunately it was all because of one book“Perception and Action in the Visual System” by Marvin Minsky and Seymour Papert, published in 1969.  This book discussed the limitations of neural networks and helped spark skepticism about the potential of AI, leading to a decrease in funding and interest in AI research during the 1970s. This period became known as the "AI winter," where enthusiasm for AI projects significantly waned due to unmet expectations and the challenges highlighted in Minsky and Papert’s work.

However, thanks to the resurgence of interest in recent years, especially after the publication of “Deep Learning” by Ian Goodfellow, Yoshua Bengio, and Aaron Courville in 2016, neural networks are back in the spotlight. This book laid the groundwork for many advancements we see today, particularly in deep learning, which has propelled machine learning into the limelight.

Types of ML

Alright do you remember that silly little training type I told you early, well guess what? There are more. 

Supervised Learning: In this type, the model is trained on labeled data, which means the input data is paired with the correct output. For example, in email spam detection, the model learns to identify spam emails based on examples that have been pre-labeled as spam or not spam.

Unsupervised Learning: Here, the model is trained on unlabeled data, meaning there are no predefined categories. Instead, it must find patterns and relationships on its own. Clustering customers into different segments based on purchasing behavior is a common application of unsupervised learning.

Reinforcement Learning: This type of learning involves agents that learn to make decisions through trial and error. They receive rewards or penalties based on their actions, enabling them to improve over time. This approach is often used in robotics and game playing, where agents learn to navigate complex environments.

Common Algorithms 

Now for this part I will recommend a little bit of abstraction (OOP reference) if you are a beginner as things can get complicated so I will just superficially explain very simply what they are

Linear Regression (You know, from your maths class?)

• What it is: Imagine trying to predict how tall someone will be based on their age. Linear regression draws a straight line through data points to find the best fit, allowing us to make predictions about future valuess
• How it works: It uses a formula to relate the input (like age) to the output (height) so that we can estimate the output for new input data

Logistic Regression 

• -What it is: Despite its name, this algorithm is used for binary classification problems, like deciding if an email is spam or not (yes or no)
• -How it works: It takes input data and applies a function to predict probabilities, giving us a score between 0 and 1. If the score is above a certain threshold, we classify it as "yes" (spam), and below that threshold, it's "no" (not spam)

Decision Trees

• What it is: Think of a flowchart for making decisions. Decision trees break down complex decision-making into simpler, yes-or-no questions
•  How it works: Starting from a root question, it branches out based on the answers, leading to a final decision. For instance, it could help determine whether to play outside based on the weather and temperature.

Neural Networks

• What it is: Inspired by the human brain, neural networks consist of layers of interconnected nodes (like neurons).
• How it works: Each node processes input data and passes it on to the next layer. It’s particularly good at recognizing patterns, such as identifying faces in photos or understanding spoken words. The more layers, the deeper the learning!

K-Means Clustering

• How it works: You decide how many groups (or clusters) you want. The algorithm then finds the best way to divide the data into those clusters based on similarity, making it great for things like customer segmentation.
• What it is: Imagine organizing a group of friends based on common interests. K-means clustering groups similar data points together.

Support Vector Machines (SVM)

• What it is: Think of drawing a line to separate two different types of apples in a basket: green and red. SVM finds the best line (or hyperplane) that separates different classes of data.
• How it works: It maximizes the distance between the line and the closest data points from each class, helping it to classify new data accurate

End

And there you have it! We’ve journeyed through the basic starter kit of machine learning, exploring its origins, how it works and the common algorithms!

So, let’s hope that wasn’t too much for you! I might have gotten a little carried away, but you know how it is when you like something a little bit too much... If you have any questions or thoughts, feel free to drop them in the comments below. Here’s to your adventure in machine learning—happy learning!

 Regina Garrido