Detection of Sparse Principal Components Based on Graph Clustering Methods

Student: Kuznetsov Kirill

Supervisor: Maxim Panov

Educational Programme: Mathematical Methods of Optimization and Stochastics (Master)

Final Grade: 7

Year of Graduation: 2017

In this paper we consider methods for searching for sparse principal components. This class of problems is necessary for lowering the dimension while preserving the interpretability of the features. Based on the existing methods of dimensionality reduction and the method of clustering graphs, a new method of dimension reduction with the use of graph clustering is proposed. The proposed algorithm uses the method of graph clustering (Louvain), by which groups of dependent attributes are distinguished. The method of principal components is applied to groups of attributes. This gives us the problem of sparse principal components. In this paper, we compare the numerical results obtained with the new method and the classical SPCA.

Full text (added May 30, 2017)

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses