Contrastive Learning

What is Contrastive Learning?

A self-supervised technique where a model learns by pulling similar images together (a dog and a cropped dog) and pushing dissimilar ones apart in embedding space.

Where did the term "Contrastive Learning" come from?

SimCLR, MoCo architectures.

How is "Contrastive Learning" used today?

Basis for powerful unsupervised vision models.

Related Terms

• Self-Supervised Learning
• embedding