A machine learning paradigm that enables models to learn new tasks or concepts from very few training examples. Few-shot learning mimics human ability to generalize from limited experience, often using meta-learning techniques to quickly adapt to new scenarios.
