Artificial Intelligence · Cambridge syllabus 6.3 · Lesson 3 of 3
Machine Learning
Machine learning is when a program can automatically adapt its own processes, its data or both. Instead of a programmer writing every rule, the program improves from the data it receives, so its output can change over time without its code being rewritten. An expert system, by contrast, uses fixed rules written by people.
By the end, you can
- Understand that machine-learning systems can adapt their processes or data.
- Distinguish machine learning from fixed rule-based systems.
- Identify suitable applications of machine learning.
Machine learning
Machine learning is when a program has the ability to automatically adapt its own processes, its data, or both. Instead of a programmer writing a rule for every case, the program changes how it works as it receives more data, so it improves at its task without its code being rewritten.
The contrast with a fixed rule-based system is the thing to understand. In an expert system a person writes every rule, and the system can never do better than those rules or handle a case they do not cover. A machine-learning program is given examples, such as thousands of emails marked spam or not spam, and adapts its own data and processes to tell them apart, including by patterns nobody wrote down.
It keeps adapting while it is used. When a user marks a message as spam that the filter missed, the filter adapts, so similar messages are caught next time. A shopping site that recommends products adapts its recommendations from what each customer views and buys.
It is used where rules are impossible or impractical to write: speech and handwriting recognition, image and face recognition, recommendations, spam filtering, fraud detection, translation and self-driving cars.
Worth remembering
- Machine learning: a program automatically adapts its own processes and/or data.
- An expert system uses fixed rules written by people; machine learning adapts from data.
- Used for spam filters, recommendations, speech and image recognition.
Worked example
An email filter is shown messages that users mark as spam. It adapts the data it uses to recognise spam, so next week it catches kinds of spam it missed this week.
Practice
Warm-up · not marked
Check your understanding
Answer in your own words, then compare with the worked answer. Nothing here is saved or scored.
Worked answer
Its own processes and/or its data, automatically, as it receives more data.
Check your method, units and explanation against it.
Marked practice
Set 1 of 4 · easiest first · Checked as you go, with the worked answer. Sign in or create a free account to save your progress.