Machine Learning Basics
Level 1 · How a machine learns
What machine learning is, and how a model learns rules from examples. What features, labels and weights are, shown with house prices.
Course contentsLesson 1 of 18
Level 1 · How a machine learns
Level 2 · Getting the data ready
Level 3 · Kinds of learning
Level 4 · Classic models
Level 5 · Neural networks
Level 6 · Better models, real tools
After this lesson you can
- Say what machine learning is
- Point to the features and the label in a table
- Explain what training changes in a model
11 mincovers 5 glossary entries
What machine learning is, and how it learns
Machine learning is a way to build AI where the computer learns rules from examples. Nobody types the rules in. Arthur Samuel made the name popular in 1959, with a program that got better at checkers by playing. Today it sorts spam, sets prices and spots fraud.
The examples are the training data. In a table, each row is one example: one house, one email, one payment. Each column the model reads is a feature, like a house's size or its distance to town. The column it learns to guess is the label, like the price the house sold for.
Inside, a model is a set of numbers called parameters. A simple price model might have just two: a base price and a price per square metre. Training starts them at any value. It nudges them, row by row, until the guesses match the labels as closely as it can.
The nudging needs a way to say how wrong a guess was. If the model says 300,000 euros and the house sold for 340,000, it is 40,000 off. Training adds up those misses over every row, then moves the parameters to shrink the total. That loop is the whole of learning.
AI is the big field, and machine learning is the part that learns from data. Deep learning is the part of it that uses networks with many layers. Every deep learning model is machine learning. Not every AI program is, because some still run on rules written by hand.
The data sets the limit. A model can only learn patterns that are in its rows. If every house in the data is in one city, its guesses for another city are poor. If the labels are wrong, it learns the wrong lesson, and learns it well.
This course covers how models learn, the main kinds, and the classic ways to build one. It stops at building. How to score a model in depth, and how to run one for real users, are subjects of their own. People often rush to the model. Most of the work, and most of the wins, are in the data.
Worked example
Emma prices flats
Emma has 2,000 past flat rentals. Each row has the size, the floor and the street, plus the rent it got.
Training finds a base rent and a price for each extra square metre. For a new 60 square metre flat, the model guesses 1,200 euros a month.
The words for this lesson5
Check yourself
What does a machine learning model learn from?
Finished this lesson?
Marking it complete also marks its 5 glossary entries learned, which updates your roadmap and your saved page. 0 are already marked.