Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Heterogeneous Graph Generation: A Hierarchical Approach using Node Feature Pooling

Hritaban Ghosh, Chen Changyu, Arunesh Sinha, Shamik Sural

TL;DR

This paper addresses the challenges in heterogeneous graph generation by employing a two phase hierarchical approach called HG2NP (Heterogeneous Graph Generation using Node Feature Pooling) and conducts extensive experiments with the well-known IMDB and DBLP datasets.

Abstract

Heterogeneous graphs are present in various domains, such as social networks, recommendation systems, and biological networks. Unlike homogeneous graphs, heterogeneous graphs consist of multiple types of nodes and edges, each representing different entities and relationships. Generating realistic heterogeneous graphs that capture the complex interactions among diverse entities is a difficult task due to several reasons. The generator has to model both the node type distribution along with the feature distribution for each node type. In this paper, we look into solving challenges in heterogeneous graph generation, by employing a two phase hierarchical structure, wherein the first phase creates a skeleton graph with node types using a prior diffusion based model and in the second phase, we use an encoder and a sampler structure as generator to assign node type specific features to the nodes. A discriminator is used to guide training of the generator and feature vectors are sampled from a node feature pool. We conduct extensive experiments with subsets of IMDB and DBLP datasets to show the effectiveness of our method and also the need for various architecture components.

Heterogeneous Graph Generation: A Hierarchical Approach using Node Feature Pooling

TL;DR

This paper addresses the challenges in heterogeneous graph generation by employing a two phase hierarchical approach called HG2NP (Heterogeneous Graph Generation using Node Feature Pooling) and conducts extensive experiments with the well-known IMDB and DBLP datasets.

Abstract

Heterogeneous graphs are present in various domains, such as social networks, recommendation systems, and biological networks. Unlike homogeneous graphs, heterogeneous graphs consist of multiple types of nodes and edges, each representing different entities and relationships. Generating realistic heterogeneous graphs that capture the complex interactions among diverse entities is a difficult task due to several reasons. The generator has to model both the node type distribution along with the feature distribution for each node type. In this paper, we look into solving challenges in heterogeneous graph generation, by employing a two phase hierarchical structure, wherein the first phase creates a skeleton graph with node types using a prior diffusion based model and in the second phase, we use an encoder and a sampler structure as generator to assign node type specific features to the nodes. A discriminator is used to guide training of the generator and feature vectors are sampled from a node feature pool. We conduct extensive experiments with subsets of IMDB and DBLP datasets to show the effectiveness of our method and also the need for various architecture components.

Paper Structure

This paper contains 20 sections, 2 theorems, 1 equation, 5 figures, 14 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work and Background
  3. Methodology
  4. Phase 1: Skeleton Graph Generation
  5. Phase 2: Heterogeneous Feature Vector Assignment
  6. Evaluation Metrics
  7. Permutation Equivariance and Invariance
  8. Experiments
  9. Limitations and Discussions
  10. Data Description
  11. Digital Bibliography & Library Project (DBLP)
  12. Internet Movie Database (IMDB)
  13. Evaluation Metrics
  14. Feature Distribution EMD
  15. Type Degree Distribution MMD
  16. ...and 5 more sections

Key Result

Lemma 1

Phase 2 generator is labeled permutation equivariant.

Figures (5)

  • Figure 1: Overall generation process for HG2NP. The skeleton graph is the output of Phase 1. Phase 2 uses a GAN type structure (see details in text and Figure \ref{['fig:generation']}).
  • Figure 2: The generation process using IMDB data as example. In Phase 1, we use DiGress to output a graph with node types, after which the first part in Phase 2 is to perform message passing to embed neighbor node type information in each node type vector.
  • Figure 3: Example of DBLP graphs from the split criteria - (author, conference and type)
  • Figure 4: Example of the IMDB graphs from the split criteria - (movie, year, language and country)
  • Figure 5: Generated Graph Samples

Theorems & Definitions (4)

  • Lemma 1
  • proof
  • Lemma 2
  • proof