Parallel Computation In Finding Near Neighbourhoods

The problem considered in this article stems from the observation that practical applications of near set theory requires efficient determination of all the tolerance classes containing objects from the union of two disjoints sets. Near set theory consists in extracting perceptually relevant information from groups of objects based on their descriptions. Tolerance classes are sets where all the pairs of objects within a set must; satisfy the tolerance relation and the set is maximal with respect to inclusion. Finding such classes is a computationally complex problem, especially in the case of large data sets or sets of objects with similar features. The contribution of this article is a parallelized algorithm for finding tolerance classes using NVIDIA's Compute Unified Device Architecture (CUDA). The parallelized algorithm is illustrated in terms of a content-based image retrieval application.