Types of Machine Learning
Machine Learning is not one single method. There are different approaches for teaching machines how to learn from data and experience.
In this lesson, we'll learn the three major types of Machine Learning: Supervised Learning, Unsupervised Learning, and Reinforcement Learning.
Supervised Learning
Learns from examples that already have known answers.
Unsupervised Learning
Finds patterns or groups in data without known answers.
Reinforcement Learning
Learns through actions, feedback, rewards, and penalties.
By the End of This Lesson
- Understand what Supervised Learning means.
- Understand what Unsupervised Learning means.
- Understand how Reinforcement Learning works.
- Compare the three main learning approaches.
- Recognize simple real-world examples of each type.
Supervised Learning
Supervised Learning is a type of Machine Learning where the system learns from examples that already have known answers.
In other words, the training data contains both the input and the expected output.
The model learns the relationship between the input and output and then uses that learned relationship to make predictions for new data.
📚 Learning With Labeled Examples
The data used in Supervised Learning is commonly called labeled data.
For example, imagine we want to build a system that identifies whether an email is spam.
We can provide examples like these:
| Label | |
|---|---|
| "You won a free prize!" | 🚨 Spam |
| "Your order has shipped." | ✅ Not Spam |
| "Claim your reward now." | 🚨 Spam |
The model can study these examples and learn patterns associated with spam and non-spam messages.
🤖 How Supervised Learning Works
📚 Labeled Training Data
│
▼
🤖 Machine Learning
│
▼
🔍 Learn Relationship
│
▼
🧠 Trained Model
│
▼
📩 New Email
│
▼
📊 Prediction
│
▼
🚨 Spam / ✅ Not Spam
🎯 What Can Supervised Learning Predict?
Supervised Learning is commonly used for two broad kinds of problems: classification and regression.
Classification means predicting a category.
For example:
📧 Email ↓ 🤖 Model ↓ 🚨 Spam OR ✅ Not Spam
Regression means predicting a numerical value.
For example, a model could predict the price of a house based on information such as its size, location, and number of bedrooms.
🏠 House Information
↓
🤖 Model
↓
💰 Predicted Price
Simple Definition
Supervised Learning is a Machine Learning approach where a model learns from labeled examples containing known inputs and outputs.
Next, we'll use a simple real-world example to make Supervised Learning even easier to understand.
A Simple Supervised Learning Example
Let's use a simple example to understand how Supervised Learning works in practice.
Imagine that we want to build a system that predicts whether a student will pass an exam based on the number of hours they studied.
📚 Training Data
First, we give the system examples where the correct result is already known.
| Study Hours | Known Result |
|---|---|
| 1 hour | ❌ Failed |
| 2 hours | ❌ Failed |
| 4 hours | ✅ Passed |
| 6 hours | ✅ Passed |
| 8 hours | ✅ Passed |
Notice that every training example contains two things:
📥 Input: Number of study hours
📤 Known Output: Passed or Failed
🤖 Learning From the Examples
📚 Labeled Examples
│
▼
🤖 Supervised Learning
│
▼
🔍 Learn Relationship
│
▼
🧠 Trained Model
The model examines the examples and learns a relationship between study time and exam results.
It may learn that students who study more hours tend to have a higher probability of passing.
🎯 Predicting a New Student
Now suppose a new student studies for 5 hours. We don't yet know whether that student will pass.
📚 New Student
│
▼
⏱️ 5 Hours
│
▼
🧠 Trained Model
│
▼
📊 Prediction
│
▼
✅ Higher Probability of Passing
The important point is that the model is making a prediction based on patterns learned from previous labeled examples.
⚠️ Prediction Is Not a Guarantee
The model does not know everything about the new student.
Other factors such as prior knowledge, exam difficulty, concentration, sleep, and many other variables could affect the final result.
Therefore, the model's prediction is an estimate, not a guaranteed outcome.
Key Idea
Supervised Learning uses examples with known answers to learn how inputs are related to outputs.
🔑 Remember the Pattern
📚 Labeled Data
↓
🤖 Learn From Examples
↓
🧠 Trained Model
↓
📥 New Data
↓
📊 Prediction
This same idea can be used for tasks such as spam detection, house-price prediction, image classification, and many other Machine Learning problems.
Next, we'll look at a different approach where the data does not come with known answers: Unsupervised Learning.
Unsupervised Learning
Unsupervised Learning is a type of Machine Learning where the system works with data that does not have predefined labels or known answers.
Instead of telling the model what the correct answer is, we give it data and ask it to discover useful patterns, relationships, or groups within that data.
🔍 No Known Answers
In Supervised Learning, the training data contains known answers. In Unsupervised Learning, those answers are not provided.
For example, imagine an online store has information about thousands of customers. The store may know things such as what customers purchased, how often they purchased, and how much they spent.
But the store may not already know which customers belong to which groups.
📊 Customer Data
│
▼
🤖 Unsupervised Learning
│
▼
🔍 Find Patterns
│
▼
📦 Discover Groups
👥 Customer Grouping
An Unsupervised Learning algorithm can analyze customer behavior and discover groups of customers with similar characteristics.
For example, it might discover groups such as:
🛍️ Frequent Shoppers
Customers who purchase products regularly.
💰 High-Spending Customers
Customers who tend to make larger purchases.
🆕 Occasional Shoppers
Customers who purchase products less frequently.
The important point is that nobody had to provide these groups as predefined labels.
The algorithm attempts to discover meaningful structure within the data.
📊 Clustering
One common Unsupervised Learning technique is called clustering.
Clustering attempts to organize similar data points into groups while separating data points that are less similar.
📊 Unlabeled Data
● ● ● ▲ ▲ ▲
● ● ● ▲ ▲ ▲
● ▲
+
■ ■ ■
■ ■ ■
↓
🤖 Clustering
↓
📦 Group 1
📦 Group 2
📦 Group 3
🎯 What Is the Goal?
The goal of Unsupervised Learning is not to predict a known answer.
Instead, the system attempts to discover structure, relationships, or patterns that are already present in the data.
Supervised Learning
📚 Data + Known Answers → Learn to Predict
Unsupervised Learning
📊 Data Without Known Answers → Discover Patterns
Next, we'll use a simple example to see how Unsupervised Learning can discover groups from data.
A Simple Unsupervised Learning Example
Let's understand Unsupervised Learning using a simple customer segmentation example.
Imagine an online store has information about its customers, including how often they purchase products and how much they usually spend.
However, the store does not already know which customers belong to which groups.
📊 Customer Data
| Customer | Purchases | Average Spending |
|---|---|---|
| Customer A | Very Often | High |
| Customer B | Often | High |
| Customer C | Rarely | Low |
| Customer D | Rarely | Low |
| Customer E | Sometimes | Medium |
🔍 Finding Groups
We don't tell the Machine Learning system what the customer groups should be.
Instead, the system analyzes the available data and looks for customers with similar behavior.
📊 Customer Data
│
▼
🤖 Unsupervised Learning
│
▼
🔍 Compare Similarities
│
▼
📦 Discover Groups
👥 Possible Groups
High-Value Customers
Customers who purchase frequently and spend more.
Regular Customers
Customers who purchase products at a moderate frequency.
Occasional Customers
Customers who purchase less frequently and spend less.
These groups are examples of what an algorithm might discover. The exact groups depend on the data and the algorithm being used.
🧠 What Makes This Unsupervised?
The key point is that we did not provide labels such as "High-Value Customer", "Regular Customer", or "Occasional Customer" before training.
The system is trying to discover structure in the data by itself.
Supervised Learning
📚 Examples + Known Labels → Learn to Predict Labels
Unsupervised Learning
📊 Data Without Labels → Discover Groups or Patterns
⚠️ The Groups Are Not Automatically Meaningful
There is an important detail beginners often miss.
An algorithm can create groups, but humans may still need to examine those groups and determine whether they are actually useful.
For example, discovering three customer groups does not automatically mean that those three groups are the best way for the business to understand its customers.
Key Idea
Unsupervised Learning looks for useful patterns or structure in data without being given predefined labels.
Next, we'll learn about the third major type of Machine Learning: Reinforcement Learning.
A Simple Reinforcement Learning Example
Let's understand Reinforcement Learning with a simple example: teaching a computer to move through a game.
Imagine a small game where an AI-controlled character needs to reach a goal while avoiding obstacles.
🎮 The Environment
The game board is the environment. The AI can observe its current position and choose an action.
┌───────┬───────┬───────┬───────┐ │ │ │ │ 🏆 │ ├───────┼───────┼───────┼───────┤ │ │ 🧱 │ │ │ ├───────┼───────┼───────┼───────┤ │ 🤖 │ │ │ │ └───────┴───────┴───────┴───────┘ 🤖 = Agent 🧱 = Obstacle 🏆 = Goal
🤖 The Agent Takes an Action
The AI might choose to move left, right, up, or down.
🤖 Agent
⬆️
⬅️ 🤖 ➡️
⬇️
Choose an Action
The result of the action determines what happens next.
🏆 Rewards and Penalties
Suppose the AI reaches the goal. The system gives it a positive reward.
🤖 ────────► 🏆
│
▼
🏆 Positive Reward
Now suppose the AI moves into an obstacle. The system could give it a penalty.
🤖 ────────► 🧱
│
▼
❌ Penalty
🔄 Learning Through Repeated Attempts
At the beginning, the AI may make many poor decisions. It does not immediately know the best path.
By repeatedly interacting with the environment, it receives feedback and gradually learns which actions tend to produce better results.
Try Action
↓
Observe Result
↓
Receive Reward / Penalty
↓
Learn From Feedback
↓
Try Again
↓
Improve Behavior
🎯 What Does the AI Learn?
Over many attempts, the system can learn that some actions are more useful in particular situations than others.
Eventually, it may discover a strategy that allows it to reach the goal more effectively.
Early Attempts
🤖 → 🧱 ❌
🤖 → 🧱 ❌
🤖 → 🏆 ✅
Later Attempts
🤖 ────────► 🏆
✅
🌍 Real-World Applications
The same basic idea can be used in more complex situations.
Games
AI agents can learn strategies by interacting with games and receiving feedback.
Robotics
Robots can learn useful behaviors through interaction with their environment.
⚠️ Important Point
Reinforcement Learning is not simply about receiving a reward after every action.
The system needs to learn which actions and strategies produce better long-term results.
Key Idea
Reinforcement Learning teaches an agent through interaction, feedback, rewards, and penalties.
Now that we've seen all three major approaches, let's compare them side by side.
Example of Reinforcement Learning
A robot learning to navigate a maze or a game-playing agent learning to improve its moves is a classic example.
The system learns from rewards when it makes good choices and penalties when it makes poor ones.
Comparing the Three Types of Machine Learning
We have now looked at Supervised Learning, Unsupervised Learning, and Reinforcement Learning.
The easiest way to understand the difference is to focus on what kind of feedback the system receives while learning.
👨🏫 Supervised Learning
The system learns from examples where the correct answer is already known.
📚 Labeled Data
↓
🤖 Learn From Examples
↓
🧠 Model
↓
📥 New Data
↓
📊 Prediction
Think: A teacher gives you examples together with the correct answers.
🔍 Unsupervised Learning
The system receives data without predefined answers and attempts to discover patterns or groups.
📊 Unlabeled Data
↓
🤖 Analyze Data
↓
🔍 Find Patterns
↓
📦 Discover Groups
Think: You receive a large collection of information and try to discover how it is organized.
🎮 Reinforcement Learning
The system learns by interacting with an environment and receiving feedback through rewards or penalties.
🤖 Agent ↓ 🎮 Action ↓ 🌍 Environment ↓ 🏆 Reward / ❌ Penalty ↓ 🧠 Learn ↓ 🔄 Try Again
Think: You learn by trying different actions and discovering which actions produce better results.
⚖️ Side-by-Side Comparison
| Type | Learning From | Main Goal | Example |
|---|---|---|---|
| 👨🏫 Supervised | Labeled data | Predict known types of outcomes | Spam detection |
| 🔍 Unsupervised | Unlabeled data | Discover patterns or groups | Customer segmentation |
| 🎮 Reinforcement | Actions and feedback | Learn better behavior over time | Game-playing agent |
🧠 An Easy Way to Remember
Supervised
Known Answers
Learn from labeled examples.
Unsupervised
Find Patterns
Discover structure in unlabeled data.
Reinforcement
Learn From Feedback
Improve actions using rewards and penalties.
Important
These three categories are useful foundations, but real Machine Learning systems can be more complicated.
Some systems can combine different techniques, and the best approach depends on the problem, the available data, and the desired outcome.
Now let's review the most important ideas from this lesson.
What You Learned
Congratulations! You have completed Lesson 10.
You now understand the three major approaches to Machine Learning and how they differ from one another.
👨🏫 Supervised Learning
Supervised Learning uses labeled examples where the correct output is already known.
The model learns the relationship between inputs and outputs and uses that knowledge to make predictions on new data.
🔍 Unsupervised Learning
Unsupervised Learning works with data that does not have predefined labels.
The system attempts to discover useful patterns, relationships, or groups within the data.
🎮 Reinforcement Learning
Reinforcement Learning involves an agent interacting with an environment.
The agent learns from actions and feedback such as rewards and penalties.
⚖️ The Main Difference
The easiest way to remember the three approaches is to focus on how the system learns.
👨🏫 Supervised
Known Answers
↓
Learn to Predict
🔍 Unsupervised
No Known Answers
↓
Discover Patterns
🎮 Reinforcement
Actions + Feedback
↓
Improve Behavior
🌍 Different Problems Need Different Approaches
There is no single Machine Learning approach that is best for every problem.
The right approach depends on the problem, the available data, and the desired outcome.
Remember It This Way
📚 Supervised
"Here are examples with answers."
🔍 Unsupervised
"Here is data. Find useful patterns."
🎮 Reinforcement
"Try actions and learn from feedback."
The most important idea from this lesson is:
Supervised Learning learns from labeled examples, Unsupervised Learning discovers patterns in unlabeled data, and Reinforcement Learning learns through actions and feedback.
In the next lesson, we'll go deeper into how a Machine Learning model is trained and how we evaluate whether it is actually learning useful patterns.