Wire before You Walk

Tesfa Asmara, Dhananjay Bhaskar, Ian Adelstein, Smita Krishnaswamy, Michael Perlmutter

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Node embeddings aim to associate a vector to every vertex of a graph which can then be used for downstream tasks such as clustering, classification, or link prediction. Many popular node embeddings such as node2vec and DeepWalk are based upon counting which nodes frequently co-occur in random walks of the graph. In this paper, we show that the performance of such algorithms can be improved by rewiring the edges of the graph through a variety of network indices before running DeepWalk. These rewirings effectively give the random walker an inductive bias and increase the accuracy of a logistic regression classifier applied to the node embedding on several benchmark data sets.

Original languageEnglish
Title of host publicationConference Record of the 57th Asilomar Conference on Signals, Systems and Computers, ACSSC 2023
EditorsMichael B. Matthews
PublisherIEEE Computer Society
Pages714-716
Number of pages3
ISBN (Electronic)9798350325744
DOIs
StatePublished - 2023
Event57th Asilomar Conference on Signals, Systems and Computers, ACSSC 2023 - Pacific Grove, United States
Duration: 29 Oct 20231 Nov 2023

Publication series

NameConference Record - Asilomar Conference on Signals, Systems and Computers
ISSN (Print)1058-6393

Conference

Conference57th Asilomar Conference on Signals, Systems and Computers, ACSSC 2023
Country/TerritoryUnited States
CityPacific Grove
Period29/10/231/11/23

Keywords

  • learning on graphs
  • node embeddings
  • skip-gram methods

Fingerprint

Dive into the research topics of 'Wire before You Walk'. Together they form a unique fingerprint.

Cite this