Metric Learning

Fact

A training method that learns how close or far examples should be.

In Plain Words

Metric Learning is a school cafeteria seating chart for data. Dog photos sit together, and weird raccoon photos get the lonely table.

You meet it in face ID and product search. It also helps recommendations keep similar things close.

Related Concepts

• Representation Learning
• Vector Embedding
• Contrastive Learning
• k-Nearest Neighbors

Representation Learning
Metric Learning is a kind of Representation Learning. It focuses on comparing examples.

Embedding
Metric Learning turns examples into Embeddings with useful distance.

Contrastive Learning
Contrastive Learning often trains Metric Learning by pulling and pushing examples.

KNN
Metric Learning gives KNN a better distance to trust.