GPEN: Global Position Encoding Network for Enhanced Subgraph Representation Learning

Nannan Wu, Yuming Huang, Yiming Zhao, Jie Chen, Wenjun Wang

Abstract:

Subgraph representation learning has attracted growing interest due to its wide applications in various domains. However, existing methods primarily focus on local neighborhood structures while overlooking the significant impact of global structural information, in particular the influence of multi-hop neighbors beyond immediate neighborhoods. This presents two key challenges: how to effectively capture the structural relationships between distant nodes, and how to prevent excessive aggregation of global structural information from weakening the discriminative ability of subgraph representations.
To address these challenges, we propose GPEN (Global Position Encoding Network). GPEN leverages a hierarchical tree structure to encode each node's global position based on its path distance to the root node, enabling a systematic way to capture relationships between distant nodes. Furthermore, we introduce a boundary-aware convolution module that selectively integrates global structural information while maintaining the unique structural patterns of each subgraph. Extensive experiments on eight public datasets identify that GPEN significantly outperforms state-of-the-art methods in subgraph representation learning.

Lay Summary: Imagine you're a detective trying to identify suspicious groups of people by only looking at who talks to whom within each group, but ignoring how these groups connect to the rest of the city—you'd miss important clues like whether they're getting money from known criminals several steps away. This is exactly the problem computers face when analyzing small sections of interconnected data, like social networks or financial transactions: they focus on local patterns but miss the bigger picture. We built a system called GPEN that gives each person a "global address" based on their position in the entire network, like noting whether someone lives in the downtown business district versus a remote suspicious area. Our approach helps computers spot the difference between groups that look similar up close but are actually very different when you consider their connections to the wider world. This breakthrough significantly improves how computers detect fraud, understand biological systems, and analyze any situation where small groups are embedded within larger networks.

Primary Area: Deep Learning->Graph Neural Networks

Keywords: Subgraph Representation Learning

Submission Number: 16318