Speeding up Relief algorithms with k-d trees

There are certain problems in machine learning which desire special attention when we scale up the size of the data or move towards data mining. One of them is the problem of searching nearest neighbours of a given point in k dimensional space. If the space is <k than k-d trees can solve the problem in asymptotically optimal time under certain conditions. We investigate the use of k-d trees in nearest neighbour search in the family of attribute estimation algorithms Relief on typical machine learning databases and examine their performance under various