Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

On the Cross-type Homophily of Heterogeneous Graphs: Understanding and Unleashing

Zhen Tao, Ziyue Qiao, Chaoqi Chen, Zhengyi Yang, Lun Du, Qingqiang Sun

TL;DR

This paper addresses the challenge of measuring and leveraging homophily in heterogeneous graphs by introducing Cross-Type Homophily Ratio (CHR), which quantifies cross-type label-relevance via target information. It then proposes Cross-Type Homophily-guided Graph Editing (CTHGE), a two-phase, CHR-driven graph editing framework that prunes semantically misaligned cross-type edges and refines the remaining connections through a target-driven auxiliary learning paradigm and iterative logits refinement. Theoretical analysis links CHR to HGNN generalization, and extensive experiments on five HG datasets with nine HGNNs show consistent improvements, up to over 25% relative gains in node classification. Overall, CHR provides a principled, scalable lens on cross-type information flow in HGs, and CTHGE offers a practical, plug-in method to boost HGNN performance across diverse architectures.

Abstract

Homophily, the tendency of similar nodes to connect, is a fundamental phenomenon in network science and a critical factor in the performance of graph neural networks (GNNs). While existing studies primarily explore homophily in homogeneous graphs, where nodes share the same type, real-world networks are often more accurately modeled as heterogeneous graphs (HGs) with diverse node types and intricate cross-type interactions. This structural diversity complicates the analysis of homophily, as traditional homophily metrics fail to account for distinct label spaces across node types. To address this limitation, we introduce the Cross-Type Homophily Ratio (CHR), a novel metric that quantifies homophily based on the similarity of target information across different node types. Additionally, we propose Cross-Type Homophily-guided Graph Editing (CTHGE), a novel method for improving heterogeneous graph neural networks (HGNNs) performance by optimizing cross-type connectivity using Cross-Type Homophily Ratio. Extensive experiments on five HG datasets with nine HGNNs validate the effectiveness of CTHGE, which delivers a maximum relative performance improvement of over 25% for HGNNs on node classification tasks, offering a fresh perspective on cross-type homophily in HGs learning.

On the Cross-type Homophily of Heterogeneous Graphs: Understanding and Unleashing

TL;DR

This paper addresses the challenge of measuring and leveraging homophily in heterogeneous graphs by introducing Cross-Type Homophily Ratio (CHR), which quantifies cross-type label-relevance via target information. It then proposes Cross-Type Homophily-guided Graph Editing (CTHGE), a two-phase, CHR-driven graph editing framework that prunes semantically misaligned cross-type edges and refines the remaining connections through a target-driven auxiliary learning paradigm and iterative logits refinement. Theoretical analysis links CHR to HGNN generalization, and extensive experiments on five HG datasets with nine HGNNs show consistent improvements, up to over 25% relative gains in node classification. Overall, CHR provides a principled, scalable lens on cross-type information flow in HGs, and CTHGE offers a practical, plug-in method to boost HGNN performance across diverse architectures.

Abstract

Homophily, the tendency of similar nodes to connect, is a fundamental phenomenon in network science and a critical factor in the performance of graph neural networks (GNNs). While existing studies primarily explore homophily in homogeneous graphs, where nodes share the same type, real-world networks are often more accurately modeled as heterogeneous graphs (HGs) with diverse node types and intricate cross-type interactions. This structural diversity complicates the analysis of homophily, as traditional homophily metrics fail to account for distinct label spaces across node types. To address this limitation, we introduce the Cross-Type Homophily Ratio (CHR), a novel metric that quantifies homophily based on the similarity of target information across different node types. Additionally, we propose Cross-Type Homophily-guided Graph Editing (CTHGE), a novel method for improving heterogeneous graph neural networks (HGNNs) performance by optimizing cross-type connectivity using Cross-Type Homophily Ratio. Extensive experiments on five HG datasets with nine HGNNs validate the effectiveness of CTHGE, which delivers a maximum relative performance improvement of over 25% for HGNNs on node classification tasks, offering a fresh perspective on cross-type homophily in HGs learning.
Paper Structure (15 sections, 1 theorem, 26 equations, 7 figures, 4 tables)

This paper contains 15 sections, 1 theorem, 26 equations, 7 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminary Concepts
  4. Understanding and Unleashing the Cross-type Homophily of HGs
  5. Understanding of Cross-Type Homophily
  6. Unleashing CHR to Boost HGNNs
  7. Optimization Perspective.
  8. Complexity Analysis
  9. Experimental Evaluation
  10. Experimental Setup
  11. Main Results
  12. Comparison with Graph Structure Learning Methods
  13. Ablation Study
  14. Hyper-parameter Study
  15. Conclusion

Key Result

theorem 1

Let $\mathcal{G} = (\mathcal{V}, \mathcal{E}, \phi, \psi)$ denote an HG. We consider a binary classification problem involving node classification with an HGNN across the entire graph $\mathcal{G}$. Using target information as the classification criterion, we model the distribution of non-target nod

Figures (7)

  • Figure 1: Graph homophily in diverse cases where colors denote node labels: (a) the homophily in homogeneous graphs is considered based on node labels; (b) in HGs, label-relevance-based methods can only assess same-type homophily while failing to measure cross-type edge; (c) our method is capable of evaluating cross-type homophily in HGs by leveraging information relevance.
  • Figure 2: Edge type distribution across different heterogeneous graphs. The proportion of Cross-type Edge Types among the edge types in HGs is significant.
  • Figure 3: Impact of CHR on HGNN Performance.
  • Figure 4: Overview of the CTHGE Method.
  • Figure 5: CHR comparison before and after CTHGE. Base CHR represents the CHR value of the HG before CTHGE, while After CTHGE CHR denotes the CHR value of the HG after applying CTHGE method.
  • ...and 2 more figures

Theorems & Definitions (5)

  • definition 1: Heterogeneous Graph
  • definition 2: Homophily Ratio in Homogeneous Graphs
  • definition 3: Node and Edge Types in Heterogeneous Graphs
  • definition 4: Cross-Type Homophily Ratio
  • theorem 1