Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SOLA-GCL: Subgraph-Oriented Learnable Augmentation Method for Graph Contrastive Learning

Tianhao Peng, Xuhong Li, Haitao Yuan, Yuchen Li, Haoyi Xiong

TL;DR

This work proposes a novel subgraph-oriented learnable augmentation method for graph contrastive learning, termed SOLA-GCL, that centers around subgraphs, taking full advantage of the subgraph information for data augmentation.

Abstract

Graph contrastive learning has emerged as a powerful technique for learning graph representations that are robust and discriminative. However, traditional approaches often neglect the critical role of subgraph structures, particularly the intra-subgraph characteristics and inter-subgraph relationships, which are crucial for generating informative and diverse contrastive pairs. These subgraph features are crucial as they vary significantly across different graph types, such as social networks where they represent communities, and biochemical networks where they symbolize molecular interactions. To address this issue, our work proposes a novel subgraph-oriented learnable augmentation method for graph contrastive learning, termed SOLA-GCL, that centers around subgraphs, taking full advantage of the subgraph information for data augmentation. Specifically, SOLA-GCL initially partitions a graph into multiple densely connected subgraphs based on their intrinsic properties. To preserve and enhance the unique characteristics inherent to subgraphs, a graph view generator optimizes augmentation strategies for each subgraph, thereby generating tailored views for graph contrastive learning. This generator uses a combination of intra-subgraph and inter-subgraph augmentation strategies, including node dropping, feature masking, intra-edge perturbation, inter-edge perturbation, and subgraph swapping. Extensive experiments have been conducted on various graph learning applications, ranging from social networks to molecules, under semi-supervised learning, unsupervised learning, and transfer learning settings to demonstrate the superiority of our proposed approach over the state-of-the-art in GCL.

SOLA-GCL: Subgraph-Oriented Learnable Augmentation Method for Graph Contrastive Learning

TL;DR

This work proposes a novel subgraph-oriented learnable augmentation method for graph contrastive learning, termed SOLA-GCL, that centers around subgraphs, taking full advantage of the subgraph information for data augmentation.

Abstract

Graph contrastive learning has emerged as a powerful technique for learning graph representations that are robust and discriminative. However, traditional approaches often neglect the critical role of subgraph structures, particularly the intra-subgraph characteristics and inter-subgraph relationships, which are crucial for generating informative and diverse contrastive pairs. These subgraph features are crucial as they vary significantly across different graph types, such as social networks where they represent communities, and biochemical networks where they symbolize molecular interactions. To address this issue, our work proposes a novel subgraph-oriented learnable augmentation method for graph contrastive learning, termed SOLA-GCL, that centers around subgraphs, taking full advantage of the subgraph information for data augmentation. Specifically, SOLA-GCL initially partitions a graph into multiple densely connected subgraphs based on their intrinsic properties. To preserve and enhance the unique characteristics inherent to subgraphs, a graph view generator optimizes augmentation strategies for each subgraph, thereby generating tailored views for graph contrastive learning. This generator uses a combination of intra-subgraph and inter-subgraph augmentation strategies, including node dropping, feature masking, intra-edge perturbation, inter-edge perturbation, and subgraph swapping. Extensive experiments have been conducted on various graph learning applications, ranging from social networks to molecules, under semi-supervised learning, unsupervised learning, and transfer learning settings to demonstrate the superiority of our proposed approach over the state-of-the-art in GCL.

Paper Structure

This paper contains 23 sections, 8 equations, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Graph Neural Networks
  4. Graph Partition Algorithm
  5. GCL with Data Augmentation
  6. Discussion on Most Relevant Works
  7. Methodology
  8. The Basic GNN Module
  9. Learnable Graph View Generator
  10. Subgraph Augmentation Selector
  11. Subgraph View Generator
  12. Subgraph View Assembler
  13. Complexity Analysis
  14. Graph Contrastive Learning in SOLA-GCL
  15. Experiments
  16. ...and 8 more sections

Figures (3)

  • Figure 1: An illustration of the proposed SOLA-GCL framework. The graph view generator is composed of three critical components: the subgraph augmentation selector, the subgraph view generator, and the subgraph view assembler. The subgraph augmentation selector learns to choose the optimal augmentation strategy for each subgraph, and the subgraph view generator outputs augmented subgraph views according to the selected strategies. The subgraph view assembler constructs an augmented graph view based on these augmented subgraph views.
  • Figure 2: We visualize the graphs in the MUTAG dataset, the deeper color indicates the more important subgraphs and the ground truth is ring structure and $\text{NO}_2$ group.
  • Figure 3: Comparisons in terms of real running time. Each model is trained for 100 epochs on each dataset and the average training time of one epoch is reported.