A fast farthest neighbor search algorithm for very high dimensional data

Computing exact nearest and farthest neighbor is a challenging task, especially in the case of high-dimensional data. Several algorithms have been proposed to tackle the nearest neighbor problem. However, not much emphasis has been given on farthest neighbor problem. Most of the previous methods addressing this problem are not efficient when applied to high-dimensional data. In this paper, we present an algorithm for exact (not approximate) farthest neighbor computation for very high dimensional data (e.g., images). Our proposed farthest neighbor algorithm is motivated by the state-of-the-art exact nearest neighbor algorithm. We consider the problem of image classification in computer vision to evaluate the performance of our proposed algorithm. For this purpose, we use CIFAR-10 dataset which includes more than 50,000 images from 10 different object categories. Experimental results show that our farthest neighbor algorithm not only achieves high accuracy but also is very fast.