Learning Through Comparison: Use Cases of Contrastive Learning

Contrastive learning is reshaping how models learn, driving widespread progress in feature learning, clustering, out-of-distribution detection, and multimodal applications. Instead of relying on rigid classifications for model training, contrastive learning helps models recognize nuanced patterns and relationships in data, making them more adaptable and better at handling real-world data. In this forum, we explore how contrastive learning works, why it’s effective, and how you can apply it in your own projects.

1. Overview of Contrastive Learning — Yin Li, PhD 01:34
2. Contrastive Learning Use Cases in Clustering and Out-Of-Distribution Detection — Chris Endemann 47:52

Slides: View the slides to grab the links referenced throughout this presentation.

Supplmental Colab notebooks: Coming soon!