Self-Supervised Learning

This is a technique where a model learns representations from data by generating its own supervision signals from the data itself. For example, a model might predict missing parts of an input or future frames in a video, creating a learning task without requiring human-labeled datasets.