V2V: Vector Embedding of a Graph and Applications

Date
2018-01-01
Authors
Nguyen, Trong
Tirthapura, Srikanta
Tirthapura, Srikanta
Major Professor
Advisor
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Altmetrics
Authors
Research Projects
Organizational Units
Computer Science
Organizational Unit
Journal Issue
Series
Department
Computer ScienceElectrical and Computer Engineering
Abstract

We present V2V, a method for embedding each vertex in a graph as a vector in a fixed dimensional space. Inspired by methods for word embedding such as word2vec, a vertex embedding is computed through enumerating random walks in the graph, and using the resulting vertex sequences to provide the context for each vertex. This embedding allows one to use well-developed techniques from machine learning to solve graph problems such as community detection, graph visualization, and vertex label prediction. We evaluate embeddings produced by V2V through comparing results obtained using V2V with results obtained through a direct application of a graph algorithm, for community detection. Our results show that V2V provides interesting trade-offs among computation time and accuracy.

Comments

This is a manuscript of a proceeding published as Nguyen, Trong Duc, and Srikanta Tirthapura. "V2V: Vector Embedding of a Graph and Applications." In 2018 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), (2018): 1175-1183. DOI: 10.1109/IPDPSW.2018.00182. Posted with permission.

Description
Keywords
Citation
DOI