kNN-Search

The k-nearest neighbor (k-NN) algorithm is an important method in statistics and machine learning. We can use it to solve classification and regression problems. In a classification setting, we first find the k closest points for a given data point using a similarity measure. Then, we determine the label of the given data point with majority vote.

This project focuses on the k-nearest neighbor search (k-NN search). Under the setting of k-NN search, we only need to find the k nearest neighbor. Further classification or regression tasks are not necessary.

We complete a CUDA implementation and a pure Cpp implementation of the plain version k-NN search. Our implementations are based on Vincent Garcia’s work. Link.

You can find the source code and sample usage from here.