Spektrální klastrování a jeho paralelizace pomocí technologie CUDA

Abstract

This paper deals with spectral clustering problem and its parallelization using CUDA technology. Paper also describes image segmentation problem, which is solved using spectral clustering. Clustering problem is outlined as well as theoretical principles and~processes of final solution. Practical solution is described in detail. Sequential and parallel solution are proposed as well as used methods and tweaks designed to decrease time and memory requirements. Paper also focuses on comparing processed solutions and~describes their differences.