This repository contains the Cython implementation of the adaptive k-nearest-neighbor algorithm [1] and the experiments from the paper.

In order to be able to run the adaptive k-NN algorithm, you will need to first compile the Cython code with the command

$ python setup.py build_ext --inplace

Afterwards, run the experiments in Adaptive k-NN Subspaces Tiny ImageNet.ipynb and generate the plots with Adaptive k-NN Figures.ipynb.

[1] D. LeJeune, R. Heckel, R. G. Baraniuk, "Adaptive Estimation for Approximate k-Nearest-Neighbor Computations," 2019. AISTATS 2019. arXiv