Aplikace metod pro redukci dimenzionality

Abstract

The master thesis deals with possible applications of selected methods for dimensionality reduction. It focuses mainly on the following methods: Singular Value Decomposition, Principal Component Analysis, Non-negative Matrix Factorization, Kernel Principal Component Analysis and CUR. These methods are used in various areas. In the thesis, basic concepts necessary for understanding of the subject matter are explained. The work continues with possible applications of selected methods and their explanation. Following part deals with experiments, which are mainly focused on an image compression, but several applications of selected methods on documents are also presented there. The aim of the thesis is to acquaint readers with the methods of dimensionality reduction, an implementation of these methods and the performance of experiments with various datasets.