British Sign Language Glossaries of Curriculum Terms

BSL AI Glossary - Nearest-Neighbour definition

Translation: Nearest Neighbour is a method that classifies new items by comparing them to the most similar examples in its training data. The training data may contain a wide range of examples for each category. For instance, dog images might include large, small, muscular, or thin dogs, while cat images might include long‑haired cats, hairless cats, or fuller cats with dense fur. Bird images might include swans, ducks, or other species, giving the system varied examples of what different types of birds look like.

When the system receives something new, such as an image of a lion, it measures how similar that image is to the stored examples. Using the Nearest Neighbour approach, it identifies the closest matches in the training data and assigns the new item to the category with the most similar examples. In this case, the lion image would most closely resemble the images in the "cat" category. Nearest Neighbour works by comparing the features of the new input with the features of the stored examples and using the closest matches to determine the most likely category.