We study the Approximate Nearest Neighbor problem for metric spaces where thequery points are constrained to lie on a subspace of low doubling dimension,while the data is high-dimensional. We show that this problem can be solvedefficiently despite the high dimensionality of the data.
展开▼