Motif-driven Subgraph Structure Learning for Graph Classification

TL;DR

This paper tackles graph classification under coarse supervision by introducing MOSGSL, a motif-driven subgraph structure learning framework. It decomposes graphs into subgraphs, learns their refined structures, and uses a motif-guided guidance module to capture discriminative, class-wise patterns, with an iterative subgraph-motif alignment mechanism. The approach demonstrates consistent improvements across five datasets and proves compatible with multiple backbones and learning procedures, highlighting its versatility as a data-centric GSL method. Overall, MOSGSL offers a principled way to leverage subgraph information and motifs to achieve more precise, personalized structure learning for graph-level tasks.

Abstract

To mitigate the suboptimal nature of graph structure, Graph Structure Learning (GSL) has emerged as a promising approach to improve graph structure and boost performance in downstream tasks. Despite the proposal of numerous GSL methods, the progresses in this field mostly concentrated on node-level tasks, while graph-level tasks (e.g., graph classification) remain largely unexplored. Notably, applying node-level GSL to graph classification is non-trivial due to the lack of find-grained guidance for intricate structure learning. Inspired by the vital role of subgraph in graph classification, in this paper we explore the potential of subgraph structure learning for graph classification by tackling the challenges of key subgraph selection and structure optimization. We propose a novel Motif-driven Subgraph Structure Learning method for Graph Classification (MOSGSL). Specifically, MOSGSL incorporates a subgraph structure learning module which can adaptively select important subgraphs. A motif-driven structure guidance module is further introduced to capture key subgraph-level structural patterns (motifs) and facilitate personalized structure learning. Extensive experiments demonstrate a significant and consistent improvement over baselines, as well as its flexibility and generalizability for various backbones and learning procedures.

Figures (8)

• Figure 1: The problem of coarse-grained supervision in graph classification.
• Figure 2: The framework of proposed MOSGSL, including a Subgraph Structure Learning Module (SGSL) and a Motif-driven Structure Guidance Module. In SGSL, we conduct structure learning on subgraphs separately and use a gate mechanism to adaptively combine the learned substructures. Important subgraphs are then filtered as candidate subgraphs, based on which motif alignment and motif extraction are iteratively conducted. Several components can be flexibly interchanged in MOSGSL.
• Figure 3: Performance of MOSGSL variants.
• Figure 4: Training dynamics.
• Figure 5: Number of motifs per class $R$.