Loading…
Glossary term
The correct answer attached to a training example, such as the category of an image.
A label is the correct answer attached to a training example: 'spam' on an email, 'cat' on a photo, or a price next to a house's details. In supervised learning the model is trained to predict labels from inputs.
Labels often come from human annotators, which makes them costly and sometimes inconsistent.
| Input | Label | |
|---|---|---|
| Example 1 | Photo of a dog | dog |
| Example 2 | Photo of a cat | cat |
| Example 3 | Photo of a dog | dog |
Inputs and their labels
emails = ["win a free prize now", "meeting at 3pm", "claim your reward"]
labels = ["spam", "not spam", "spam"] # one label per exampleLabel quality sets a ceiling on model quality, and much of the cost of building AI systems goes into producing good labels.