Supervised Learning
The model learns from examples where the correct answer is already known.
Example: Predict house pricesWhat Are the Types of Machine Learning?
Machine Learning is not one single technique. Different types of Machine Learning are used depending on what information we have and what we want the computer to learn.
The type of Machine Learning depends on how the computer learns from data.
Sometimes we give the model examples with known answers. Sometimes we give it data without answers and ask it to discover patterns. In other situations, an agent learns by interacting with an environment and receiving rewards or penalties.
The Three Main Types
At a high level, Machine Learning is commonly divided into three major categories.
Unsupervised Learning
The model receives data without known answers and looks for useful patterns or groups.
Example: Group similar customersReinforcement Learning
An agent learns by taking actions and receiving rewards or penalties from an environment.
Example: Learn to play a gameSupervised Learning
In supervised learning, the training data contains examples where the correct answer is known.
INPUT
Features
Information about the example.
→
MODEL
Learn
Finds relationships in the data.
→
KNOWN ANSWER
Label
The answer the model learns to predict.
For example, we can give a model information about houses together with their actual selling prices. The model can learn a relationship between the house information and its price.
Unsupervised Learning
In unsupervised learning, the data does not contain a known target answer for every example.
Instead, the algorithm tries to discover structure, relationships, or groups inside the data.
CUSTOMER DATA
Age • Spending • Purchase Frequency
↓
Find Similar Groups
For example, a company might have thousands of customers but no predefined customer groups. An unsupervised learning algorithm could help find groups of customers with similar behavior.
Reinforcement Learning
Reinforcement learning works differently from the previous two approaches.
Instead of learning from a fixed dataset of labelled examples, an agent interacts with an environment, takes actions, and receives feedback.
AGENT
Takes an Action
→
ENVIRONMENT
Responds
→
FEEDBACK
Reward / Penalty
Over time, the agent learns which actions are more useful for achieving its goal.
The Key Difference
SUPERVISED
Learn From Answers
Training examples include known labels.
UNSUPERVISED
Find Patterns
Training data does not provide known labels.
REINFORCEMENT
Learn From Feedback
Actions produce rewards or penalties.
Supervised Learning Has Two Important Problems
Two of the most common supervised learning problems are classification and regression.
CLASSIFICATION
Predict a Category
Example: Is an email spam or not spam?
Output → Category
REGRESSION
Predict a Number
Example: What will this house cost?
Output → NumberUnsupervised Learning Includes Clustering
One common unsupervised learning task is clustering.
Clustering attempts to organize data points into groups based on similarities between them.
Group A
Similar Customers
Group B
Similar Customers
Group C
Similar Customers
The important point is that we did not provide the algorithm with predefined customer labels. The algorithm discovers groups based on the data.
How Do We Choose the Type?
Start by asking what information you have and what you want the system to learn.
Question
Likely Approach
Do we have known answers?
Supervised Learning
Do we want to predict a category?
Classification
Do we want to predict a number?
Regression
Do we want to discover groups?
Clustering
Does an agent learn through actions and feedback?
Reinforcement Learning
These categories are not just different names for the same thing. The learning setup changes depending on whether the model receives known answers, discovers patterns, or learns through interaction and feedback.
The Learning Setup Determines the Type of ML.
Supervised learning learns from labelled examples. Unsupervised learning searches for structure in unlabelled data. Reinforcement learning learns through actions and feedback.
Which Type Would You Choose?
Imagine you have thousands of customer records and want to automatically discover groups of customers with similar purchasing behavior.
Answer
This is an unsupervised learning problem because there are no predefined customer group labels.
A clustering algorithm could be used to discover groups with similar behavior.