Graph Embedding: A Methodological Survey

Joseph R. Barr, Peter Shaw, Faisal N. Abu-Khzam, Tyler Thatcher, Toby Dylan Hocking

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Embedding a high dimensional combinatorial object like tokens in text or nodes in graphs into a lower dimensional Euclidean space is a form of (lossy) data compression. We will demonstrate a class of procedures to embed vertices of a (connected) graph into a low-dimensional Euclidean space. We explore two kinds of embedding, one node2vec, similar to word2vec, which deploys a shallow network and a recurrent network which remembers past moves and takes [sic] spatial correlations into an account. We also explore the extent in which graph embedding preserves information and the practicality of using the information stored in a compressed form to discern meaningful patterns. With growth in their popularity, we too make an extensive use of the neural networks computational frameworks; we propose the usage of various neural network architectures to implement an encoder-decoder scheme to learn 'hidden' features. Since training a network requires data, we describe various sampling techniques including novel methods to sample from a graph; one using a vertex cover and another is an Eulerian tour of a (possibly) modified graph.

Original languageEnglish (US)
Title of host publicationProceedings - 2022 4th International Conference on Transdisciplinary AI, TransAI 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages142-148
Number of pages7
ISBN (Electronic)9781665471848
DOIs
StatePublished - 2022
Event4th International Conference on Transdisciplinary AI, TransAI 2022 -
Duration: Jan 1 2022 → …

Publication series

NameProceedings - 2022 4th International Conference on Transdisciplinary AI, TransAI 2022

Conference

Conference4th International Conference on Transdisciplinary AI, TransAI 2022
Period1/1/22 → …

Keywords

  • Cluster Editing
  • Data Compression
  • Encode-Decoder
  • Graph Embedding
  • Node2vec
  • Sampling
  • Vertex Cover

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Science Applications
  • Modeling and Simulation

Fingerprint

Dive into the research topics of 'Graph Embedding: A Methodological Survey'. Together they form a unique fingerprint.

Cite this