Nearest-Neighbor Lookup in kd-Trees
Now that you have a better geometric intuition for the kd-tree, we can talk about the most interesting
operation on the kd-tree: nearest-neighbor lookup. This query works as follows: given a kd-tree
and a point in space (called the test point), which point in the kd-tree is closest to the test point?
(The point in the data set closest to the test point is called its nearest neighbor). Before we discuss
the actual algorithm for doing nearest-neighbor lookup, we'll discuss the intuition behind the algorithm.
Suppose that we have a guess of what we think the nearest neighbor to the test point is. For
example, suppose that the test point is indicated by the star and that we think the nearest neighbor
is the point connected to the star by the dashed line