Three peeks, all you need.
Few-shot learning is an AI technique where a model learns to recognize new concepts from just a handful of examples, like a child learning a new animal from a few photos.
Imagine you show a friend three pictures of a new kind of fruit, and then they can spot it in any fruit bowl. That's few-shot learning — teaching a computer to learn a new thing from just a few examples, instead of needing thousands. It's like learning to identify a 'zibble' from just three zibble photos.
This matters because collecting massive labeled datasets is expensive and impractical for rare or new tasks. You encounter it in voice assistants that learn your voice from a few phrases, or in photo apps that recognize your pet from just a few tagged pictures.
A common myth is that few-shot learning means the model learns from scratch with only a few examples. Actually, the model is pre-trained on many similar tasks, so it already knows how to learn — the few examples just fine-tune it to a new specific concept.
Few-shot learning is a subfield of machine learning where a model is trained to generalize from a limited number of labeled examples per class, typically 1 to 5. It leverages prior knowledge from a large, diverse dataset of related tasks, often using meta-learning or metric-based approaches to quickly adapt to new classes with minimal data.