Decoder. plain-English AI glossary

Feature

● Core

A single measurable trait the model uses as a clue — one ingredient in its decision.

Think of it like

The tells a doctor reads: temperature, pulse, symptoms. Each is a feature; together they point to a diagnosis.

Example

In old-school ML you’d hand-pick features like “word count” or “has a link”; deep learning learns its own instead.

How it actually works

A feature is one dimension of information about an input. Classical ML lived or died on humans engineering good features. Deep learning’s leap was learning features automatically from raw data, layer by layer, so the useful ones emerge without being named.

For product teams

Where features come from — hand-built vs learned — is the line between old ML and modern AI.

For engineers

An input variable or a learned activation dimension; the axes along which a model discriminates.

Deep learning — Learned automatically in deep learning.
Representation — Bundled into a representation.

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open Decoder Add to Chrome