Využití HPC při srovnání algoritmů pro výpočet vývoje tématu

Abstract

In this world of the Internet there exists huge collections of the text documents and it is necessary to effectively make search in them. We have to use another approach to find topic development, because we would like increase number of the relevant files and decrease number of the irrelevant files in response to a query. A proper solution would be using Fiedler vector, which has a wide using. Example of the using of Fiedler vector is calculating approximate Fiedler vector for the huge sparse matrixes. The time of this calculation rises exponentially according to dimension of the sparse matrix. The goals of the thesis are evaluating propriety of using Fiedler vector for getting topical development, optimizing and paralleling its algorithm with using HPC cluster.