Dissertation/Thesis Abstract

Feature Extraction from Network Data
by Oles, Vladyslav, Ph.D., Washington State University, 2019, 92; 27547130
Abstract (Summary)

This work explores different approaches to feature extraction from network data. The first part focuses on Boolean networks, a simplistic discrete dynamical system built over graphs. We propose a method for statistical analysis of attractors of a Boolean network, and use it to learn about its behavior patterns at the system level. The analysis aggregates information about the direct interactions between the modeled objects, encoded in a Boolean network model. In the second part, we regard graph matching, a fundamental problem in the field of machine intelligence. We introduce an approach to analyze graph data based on its representation as a metric space. To measure shape difference in graphs, we develop and implement a polynomial-time algorithm for estimating one of the Gromov–Hausdorff distances.

Indexing (document details)
Advisor: Panchenko, Alexander
Commitee: Vixie, Kevin, Cooper, Kevin
School: Washington State University
Department: Mathematics
School Location: United States -- Washington
Source: DAI-B 81/9(E), Dissertation Abstracts International
Source Type: DISSERTATION
Subjects: Applied Mathematics
Keywords: Anomaly detection, Boolean networks, Computational geometry, Graph matching, Gromov–Hausdorff distances, Shape analysis
Publication Number: 27547130
ISBN: 9781658444248
Copyright © 2020 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest