Few-shot and zero-shot learning paradigms

Machine learning paradigms that train or evaluate models to classify or recognize categories using very few or no labeled examples, relying on transferred representations or semantic descriptions.

AI-generated definition