To overcome the curse of dimensionality, approximated nearest neighbor (ANN) solutions are commonly used. In particular, a c-ANN is a solution where the distance of the retrieved point from q is at most c times the true distance from the nearest point. For the ANN problem, probabilistic dimensionality reduction such as locality sensitive hashing (LSH) [33] was proven to be useful, with query time sub-linear in n but linear in d. For very-high dimensional space this may still pose a problem [35]. Note also that the solution provided by LSH is correct only with high probability.