Hello everyone, I took the last few days to finish a package I started many months ago. It is a package written in pure Julia for nearest neighbor searches in arbitrary dimensions by the use of either kd trees or ball trees (metric trees). The ball trees can use any metric and the kd trees can use one of the Minkowski metrics.
There are currently no extensive benchmarks yet but this gist: https://gist.github.com/KristofferC/ba95bb6d9c04489bbed9 shows NearestNeighbors.jl beating the cKDTree in scipy version 0.15.1 with a factor of around 4 for knn searches with k = 5 for a tree with 100 000 points in 4 dimensions. I submitted a pull request to register it in METADATA but until then Pkg.clone should work. The link to the repo is https://github.com/KristofferC/NearestNeighbors.jl Next step would be to add multithreading support but maybe that will have to wait until the thread support in base gets more mature and I have more time :) Thank you for your time, // Kristoffer