Theory of Intelligence
MACHINE LEARNING•Supervised Learning•Classification•1.2 Multiclass Classification

Introduction

Classify into more than two categories, using binary classifiers with one-vs-rest or one-vs-one strategies, or models that handle many classes natively.

Introduction

Our 5-detector answers only "is it a 5?". But to read handwriting we need to know which digit it is — one of ten classes. That's multiclass classification (also called multinomial classification).

Everyday examples:

  • Which animal is in this photo — cat, dog, rabbit or bird?
  • Which language is this sentence written in?
  • Which department should this customer email go to — billing, technical, sales?

Some algorithms can handle many classes directly. Others are binary only, but we can still use them for multiclass problems by combining several binary classifiers cleverly. This topic covers both.

Discuss theoretical concepts, mathematical formulas, and architectural foundations.

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