Nice! Note that KD-trees *do* make sense in very large dimensional spaces, where they are used for Approximate Nearest Neighbour search. This is frequent in Computer Vision, see that project for instance: https://github.com/philbinj/fastann -- Pierre Chapuis