Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Disentangled Hyperbolic Representation Learning for Heterogeneous Graphs

Qijie Bai, Changli Nie, Haiwei Zhang, Zhicheng Dou, Xiaojie Yuan

TL;DR

The results demonstrate its superiority over state-of-the-art methods, showcasing the effectiveness of the Disentangled Hyperbolic Heterogeneous Graph Convolutional Network method in disentangling and representing heterogeneous graph data in hyperbolic spaces.

Abstract

Heterogeneous graphs have attracted a lot of research interests recently due to the success for representing complex real-world systems. However, existing methods have two pain points in embedding them into low-dimensional spaces: the mixing of structural and semantic information, and the distributional mismatch between data and embedding spaces. These two challenges require representation methods to consider the global and partial data distributions while unmixing the information. Therefore, in this paper, we propose $\text{Dis-H}^2\text{GCN}$, a Disentangled Hyperbolic Heterogeneous Graph Convolutional Network. On the one hand, we leverage the mutual information minimization and discrimination maximization constraints to disentangle the semantic features from comprehensively learned representations by independent message propagation for each edge type, away from the pure structural features. On the other hand, the entire model is constructed upon the hyperbolic geometry to narrow the gap between data distributions and representing spaces. We evaluate our proposed $\text{Dis-H}^2\text{GCN}$ on five real-world heterogeneous graph datasets across two downstream tasks: node classification and link prediction. The results demonstrate its superiority over state-of-the-art methods, showcasing the effectiveness of our method in disentangling and representing heterogeneous graph data in hyperbolic spaces.

Disentangled Hyperbolic Representation Learning for Heterogeneous Graphs

TL;DR

The results demonstrate its superiority over state-of-the-art methods, showcasing the effectiveness of the Disentangled Hyperbolic Heterogeneous Graph Convolutional Network method in disentangling and representing heterogeneous graph data in hyperbolic spaces.

Abstract

Heterogeneous graphs have attracted a lot of research interests recently due to the success for representing complex real-world systems. However, existing methods have two pain points in embedding them into low-dimensional spaces: the mixing of structural and semantic information, and the distributional mismatch between data and embedding spaces. These two challenges require representation methods to consider the global and partial data distributions while unmixing the information. Therefore, in this paper, we propose , a Disentangled Hyperbolic Heterogeneous Graph Convolutional Network. On the one hand, we leverage the mutual information minimization and discrimination maximization constraints to disentangle the semantic features from comprehensively learned representations by independent message propagation for each edge type, away from the pure structural features. On the other hand, the entire model is constructed upon the hyperbolic geometry to narrow the gap between data distributions and representing spaces. We evaluate our proposed on five real-world heterogeneous graph datasets across two downstream tasks: node classification and link prediction. The results demonstrate its superiority over state-of-the-art methods, showcasing the effectiveness of our method in disentangling and representing heterogeneous graph data in hyperbolic spaces.
Paper Structure (31 sections, 29 equations, 7 figures, 7 tables)

This paper contains 31 sections, 29 equations, 7 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Heterogeneous Graph Representation Learning
  4. Disentangling Representation Learning
  5. Hyperbolic Graph Representation Learning
  6. Preliminaries
  7. Problem Definition
  8. Hyperbolic Geometry
  9. Methodology
  10. Structural Information Learning Module
  11. Heterogeneous Graph Contrastive Learning Module
  12. Heterogeneous Hyperbolic Graph Convolutional Networks
  13. Contrastive Learning
  14. Disentangling Module
  15. Overall Framework and Model Training
  16. ...and 16 more sections

Figures (7)

  • Figure 1: (a) Homogeneous graphs have only structural information which refers to the connections between node pairs. (b) Heterogeneous graphs have the fusion information of structure and semantics which refers to the types of nodes and edges.
  • Figure 2: (a) The hierarchical structure of graphs. (b) Embedding the graph into 2-D Euclidean space. The red nodes refer to the conflicts. (c) Embedding the graph into 2-D hyperbolic space (Poincaré disk model). The measures of distance between all node pairs connected with each other are equal. (d) The distribution of the relation Author$\longleftrightarrow$Paper in dataset DBLP. (e) The distribution of the relation User$\stackrel{click}{\longleftrightarrow}$Item in dataset Alibaba. Best viewed in color.
  • Figure 3: The overall framework of $\text{Dis-H}^2\text{GCN}$. Three main modules are: (a) structural information learning module, (b) heterogeneous graph contrastive learning module and (c) disentangling module.
  • Figure 4: An example of he-HGCN$^{\mathbb{B}}$ architecture upon the toy academic network. Best viewed in color.
  • Figure 5: Mutual information of our model and -w/o disentangling on both downstream tasks.
  • ...and 2 more figures

Theorems & Definitions (3)

  • Definition 1: Heterogeneous Graphs
  • Definition 2: Heterogeneous Graph Representation Learning
  • Definition 3: Disentangled Heterogeneous Graph Representation Learning