Machine Learning Foundations · Chapter 1 · ~15 min · Beginner
What is Machine Learning? Teaching Computers from Examples
Strip away the hype around artificial intelligence. Learn how machine learning flips traditional programming on its head, discover the 3 main families of ML, and understand the difference between Classification and Regression.
- Learn how ML differs from traditional rule-based coding
- Understand Supervised, Unsupervised, and Reinforcement Learning
- Master Classification (categories) vs Regression (numbers)
- Define Features (X) and Labels (y) with clear examples
1The Big Shift: Programming vs. Machine Learning
For decades, computer programming worked in one way: a human engineer sat down and wrote out every single rule by hand:
Rules + Data = Answers
You write: if (email.contains("free money")) return SPAM;. But spammers change words to "free m0ney", and your manual rule breaks!
Data + Answers = Rules!
You give the computer 1,000,000 spam emails and 1,000,000 normal emails. The computer learns the patterns automatically!
What is the fundamental difference between traditional programming and machine learning?
2The 3 Main Families of Machine Learning
Virtually every AI and ML application fits into one of these three broad categories:
Learning with a Teacher
Trained on labeled examples (e.g. 5,000 house photos with their exact sale prices). The algorithm learns to map inputs to outputs.
Finding Hidden Patterns
No labels provided! The computer groups similar items together (e.g. clustering bank customers into spending habits).
Trial and Error
An agent interacts with an environment and learns by getting rewarded for good moves and penalized for bad ones (e.g. Chess AI, robot walking).
If a hospital feeds an algorithm 50,000 X-ray scans that are each labeled 'Pneumonia' or 'Healthy', which type of ML is being used?
3Classification vs. Regression
Within Supervised Learning, there are two main types of problems:
Classification (Predicting Categories)
The output is a discrete category or label:
- • Spam or Not Spam?
- • Fraudulent transaction or Legitimate?
- • Cat, Dog, or Bird?
Regression (Predicting Numbers)
The output is a continuous numerical value:
- • What will this house sell for? ($385,000)
- • Tomorrow's temperature in London? (68°F)
- • Uber ride arrival time? (14 minutes)
Predicting whether a bank loan applicant will 'Default' or 'Pay On Time' is an example of what type of problem?
4Features (X) and Labels (y)
In machine learning literature and code, you will always see the symbols X and y:
- Features (X): The inputs. The measurable clues the computer examines. For house prices: square footage, number of bedrooms, zip code.
- Label (y): The target output. The answer the computer is trying to predict: the sale price.
5Hands-On: Test an ML Fruit Classifier
Use the interactive model below. Adjust the weight and texture features to see how the classifier updates its prediction and confidence:
Live Machine Learning Classifier (Fruit Predictor)
Apple 🍏
P(Apple) = 95%
P(Orange) = 5%
6Chapter Knowledge Challenge
Test your machine learning intuition on classification, regression, features, and labels:
Classifying incoming emails as 'Spam' or 'Inbox' is an example of what type of machine learning task?
In a house price prediction model with square footage, number of bedrooms, zip code, and selling price, which is the Label (y)?
7Recap
- Machine Learning trains algorithms on data so computers discover patterns and rules automatically.
- The 3 families are Supervised (labeled data), Unsupervised (unlabeled clustering), and Reinforcement (reward signals).
- Classification predicts discrete categories; Regression predicts continuous numerical quantities.
- Features (X) are the input clues; Labels (y) are the target answers.
Ready for Chapter 2? Mark this chapter as complete to proceed: