Publication Date

Spring 2024

Degree Type

Master's Project

Degree Name

Master of Science in Computer Science (MSCS)

Department

Computer Science

First Advisor

Katerina Potika

Second Advisor

Navrati Saxena

Third Advisor

Robert Chun

Keywords

Community Detection, Deep Learning

Abstract

Community detection in networks is essential for understanding the complex structures of connected systems. Traditional deep learning (DL) methods such as Graph Neural Networks (GNNs) and Graph Convolutional Networks (GCNs) have shown promised results in supervised tasks, like classification, but often fail in unsupervised tasks like community detection because of the lack of labels. Self- supervised approaches where we integrate crucial community information offer a solution. This project seeks to explore DL methods for community detection, focusing specifically on using Graph Variational Autoencoders (VGAEs). While classical approaches can efficiently handle small to medium-sized networks, they typically struggle with larger-sized structures. To address these challenges, we introduce the Enhanced Community Detection with Structural Information (VGAE-ECF) method. This method enhances the input of VGAE by incorporating, not only the adjacency matrix and node features but also, information on community structures and the strengths of node relationships as reflected by edge weights. We utilize the Leiden

algorithm to identify community structures, and the K-truss algorithm, which calcu- lates the weight of an edge based on the number of triangles it participates in. Further enhancing our approach, the latent space embeddings generated by our method are then passed to a K-means clustering algorithm. This step is crucial for detecting the final community structures, effectively bridging the gap between the rich, latent representations learned by the VGAE and the community partitions.

Download

Available for download on Sunday, May 25, 2025

Share

COinS