Accelarating the Euclidean distance matrix computation using GPUs

Euclidean distance matrix is used to find out the similarity of two matrices. It is a very important matrix operation used frequently in mathematical and pattern recognition based problems. The Euclidean distance matrix is a compute intensive algorithm which is also a very important step in a majority of image/speech processing algorithms. In addition to its extensive use in signal processing, it is also used in a lot of scientific calculations. Since most of the applications deal with huge matrices, the calculation often takes a significant amount of time thus slowing down algorithms. This makes the algorithms almost impossible to be implemented for real time applications. In this paper we attempt at reducing the execution time using Graphical Processing Units (GPUs). GPUs are essentially graphics cards that available at a very affordable cost and are increasing present in all computers. We use Compute Unified Device Architecture CUDA, introduced by NVIDIA for programming the GPUs.