https://hgpu.org/?p=3800
Design and evaluation of a parallel k-nearest neighbor algorithm on CUDA-enabled GPU