Machine LearningBeginner1:22 video6 chapters

What Is Machine Learning? — lecture notes

Instead of writing rules, show the computer examples. The core idea of machine learning, its three main types and the standard workflow.

▶ Watch the animated lecture

0:001. Introduction

Introduction — What Is Machine Learning?

Traditional programs follow rules that a programmer writes. Machine learning flips this around: we show the computer examples, and it works out the rules itself.

0:112. Rules vs examples

Rules vs examples — What Is Machine Learning?

In traditional programming, you give the computer data and rules, and get answers. In machine learning, you give it data and the correct answers, and it produces the rules, which we call a model.

0:263. Types of learning

Types of learning — What Is Machine Learning?

There are three main types. Supervised learning uses labelled examples. Unsupervised learning finds structure without labels. Reinforcement learning learns by trial and error, guided by rewards.

0:384. A first example

A first example — What Is Machine Learning?

Here is supervised learning in action. Each dot is a student: hours studied and exam score. The model starts with a bad guess and adjusts itself until its line fits the data well. That adjusting process is called training.

0:555. Workflow

Real projects follow a workflow. Collect data. Prepare features. Train the model. Evaluate it on data it has never seen. Then deploy it and keep monitoring, because the world changes.

1:086. Recap

To recap. Machine learning learns rules from examples. There are three main types. Training adjusts the model to reduce its errors. And we always judge a model on data it has never seen.

Key takeaways

Machine learning learns rules (a model) from data and answers.
Supervised, unsupervised and reinforcement learning are the main types.
Training adjusts parameters to reduce errors on examples.
Models must be evaluated on data they have not seen.

Check yourself

In supervised learning the training data includes…
Show answer
Inputs and correct answers (labels) — Supervised learning learns from labelled examples.
Grouping customers into segments without labels is…
Show answer
Unsupervised learning — Finding structure without labels is unsupervised.
Why evaluate on unseen data?
Show answer
To check the model generalises instead of memorising — Only unseen data shows real-world performance.

Go deeper

What Is Machine Learning? The Learning Problem Formalised · The AI Lecture Hall
Types of Machine Learning: Supervised, Unsupervised, Self-Supervised and Reinforcement · The AI Lecture Hall
The Machine Learning Workflow: From Problem to Deployed Model · The AI Lecture Hall