# Nearest-neighbor Matching

One of the most straight-forward ways to classify documents making a rote memory of the training set $T$, and retrieving those documents from $T$ that are most similar to a new document to be classified [Cost93] [Larkey96] [Yang97] [Larkey98b] . This corresponds to using the $|d \cdot d|$ similarity metric discussed in Section (FOAref) . A weighted sum of the $k$ most similar documents' classifications can be used to pick the most highly weighted classification.

FOA © R. K. Belew - 00-09-21