Artificial Intelligence

AI Learning Types - Supervised and Unsupervised Learning analogies

Analogies are often a good way to understand AI terms. Here are some analogies to help explain Supervised and Unsupervised Learning.

Supervised and Unsupervised Learning

Analogies are often a good way to understand AI terms. Here are some analogies to help explain Supervised and Unsupervised Learning.

Our articles on Supervised Learning and Unsupervised Learning will explain these terms in more detail

Unsupervised Learning by Humans Analogy

Imagine you are given a huge box of books. You begin to sort them into groups to put on your bookshelves.

  • You cluster all the mystery novels together because you're drawn to thrillers.
  • You set aside textbooks as they're primarily for study, not leisure.
  • Comic books are grouped for their light-hearted content.

Importantly, you categorised these books based on your own set of rules, without any explicit guidance.

Unsupervised Learning by Computers Analogy

Now, suppose we present a computer with a mix of images: some are paintings, others photographs, and some sketches, without any labels to distinguish them. We then challenge the computer to organise them.

During its examination, the computer observes colour schemes, the presence of realistic elements versus artistic strokes, and the inclusion of text. It's distinguishing among them based on visual clues, not names.

In its analysis, the computer might identify:

  • Some images feature vivid colours and abstract shapes (paintings).
  • Others depict life-like details and textures (photographs).
  • A selection showcases simple lines and colours (sketches).

It categorises these images without direct instruction.

Unsupervised Machine Learning

Ultimately, while the computer might not label each image accurately as a painting, photograph, or sketch, it can group them based on the similarities and differences it detected.

Supervised Machine Learning

This is in contrast to an approach where we initially show the computer images already sorted into paintings, photographs, and sketches. There, we provided a clear framework for recognition, akin to teaching a student with specific examples.

This scenario exemplifies Supervised Learning in Machine Learning, where the machine is trained using labelled data.

Conversely, when the computer is tasked with analysing and organising the images without predefined labels, relying solely on its observations, we delve into the realm of Unsupervised Learning. This distinction underscores the difference between instructing a machine with clear guidance and allowing it to discover patterns in data independently.

March 28, 2024

Read our latest

Blog posts