Self-Supervised Learning

Overview

A machine learning approach where a model learns from unlabeled data by generating its own supervision signals. Unlike supervised learning, which requires labeled data, self-supervised learning models create a prediction task based on the data's inherent structure or context, effectively generating its own "labels." This is useful when obtaining annotated data is difficult or expensive, as models learn from the data itself.

What is Self-Supervised Learning?

A learning approach where models create their own supervision signals.

How does it work?

Generates prediction tasks from data's inherent structure.

Applications?

Pre-training language models, visual representations, speech patterns.

Related Concepts

• Machine Learning
• Supervised Learning
• Unsupervised Learning
• Deep Learning