Learning the structure of manifolds using random projections

Abstract

We present a simple variant of the k-d tree which automatically adapts to intrinsic low dimensional structure in data. 1