The following code is a parallel kNN implementation that uses GPUs for the high dimensional data in text classification. You can use it to classify documents using kNN or to generate meta-features based on the distances between a query document and its k nearest neigbors