Theory of Intelligence
MACHINE LEARNING•Supervised Learning•Classification•1.3 Multilabel & Multioutput

Introduction

When one example can belong to several classes at once (multilabel), and when a model must predict many multiclass values at once (multioutput).

Introduction

So far, every example belonged to exactly one class: a digit is a 5 or a 3, never both. But many real problems aren't like that:

  • A photo can contain a dog and a cat and a person at the same time.
  • A news article can be about politics and economics.
  • A film can be tagged comedy, romance and drama.

This topic covers two more general kinds of classification:

TypeEach example has…Example
binaryone of 2 classesspam / not spam
multiclassone of many classeswhich digit
multilabelseveral yes/no labels at oncewhich people appear in a photo
multioutputseveral labels, each with many classesthe colour of every pixel in a cleaned-up image

Discuss theoretical concepts, mathematical formulas, and architectural foundations.

Supports inline code with `backticks`
No comments yet. Be the first to start the discussion!