Constrained Directed Graph Clustering through Differential Equations Open Access

Zhou, Ziyu (Spring 2023)

Permanent URL: https://etd.library.emory.edu/concern/etds/8910jv96r?locale=en
Published

Abstract

This thesis introduces an algorithmic approach to the problem of clustering a weighted directed graph under constraints such as cardinality or membership requirements. The proposed approach extends a two-level iterative approach for solving weighted undirected graph minimum-cut problems to the minimum-cut and general clustering problems of directed graphs. This two-level method restates the constrained minimum-cut problems as matrix nearness problems, for which the key to numerical solutions is the gradient systems of matrix differential equations for minimizing a functional of an eigenvalue and eigenvectors of the graph Laplacians. In the proposed approach, a directed graph is transformed into an undirected one that preserves information about directionality. Numerical experiments are presented to validate the proposed approach and compare it with existing ones for unconstrained cases. The major contribution of this work is the ability to adapt a versatile technique to directed graphs and further requirements on clustering criteria.

Table of Contents

Chapter 1 Introduction

Chapter 2 Graphs background

Chapter 3 Minimum-cut and clustering methods for directed graphs

Chapter 4 Numerical experiments

Chapter 5 Conclusions

About this Honors Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
School
Department
Degree
Submission
Language
  • English
Research Field
Keyword
Committee Chair / Thesis Advisor
Committee Members
Last modified

Primary PDF

Supplemental Files