WalkRNN: Reading Stories From Property Graphs

19 Oct 2019

Co-Author: Deborah Tylor Tylor Data Services, LLC dtylor@tylordata.com

Co-Author: Mirco A. Mannucci HoloMathics, LLC mirco@holomathics.com

Co-Author: Joseph Haaga Georgia Institute of Technology jhaaga3@gatech.edu

Abstract WalkRNN, the approach described herein, leverages research in learning continuous representations for nodes in networks, layers in features captured in property graph attributes and labels, and uses Deep Learning language modeling via Recurrent Neural Networks to read the grammar of an enriched property graph. We then demonstrate translating this learned graph literacy into actionable knowledge through graph classification tasks.

Keywords Deep Learning * Graph Mining * Random Walk * Language Model * Graph Language * RNN * GraphWave

Read WalkRNN: Reading Stories From Property Graphs

Published online at viXra