Iteratively Refined Early Interaction Alignment for Subgraph Matching based Graph Retrieval

2025-10-26

AI Summary

This paper addresses the limitation in subgraph matching—suboptimal early inter-graph alignment leading to degraded graph retrieval performance—by proposing IsoNet++, an iterative early-interaction alignment framework. Methodologically, IsoNet++ introduces three key innovations: (1) a partner interaction mechanism over cross-graph node pairs, explicitly modeling structural correspondences between graphs; (2) inter-layer cross-graph message passing coupled with a dynamic lazy-update alignment strategy to mitigate error accumulation; and (3) hierarchical GNN-based iterative reconstruction guided by edge existence discrepancies, enhancing structural sensitivity. Evaluated on multiple standard benchmarks, IsoNet++ consistently outperforms state-of-the-art methods. Ablation studies confirm that each component contributes robust and consistent accuracy gains. Overall, IsoNet++ establishes a more robust and interpretable alignment paradigm for subgraph-isomorphism-driven graph retrieval.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Graph retrieval based on subgraph isomorphism has several real-world applications such as scene graph retrieval, molecular fingerprint detection and circuit design. Roy et al. [35] proposed IsoNet, a late interaction model for subgraph matching, which first computes the node and edge embeddings of each graph independently of paired graph and then computes a trainable alignment map. Here, we present IsoNet++, an early interaction graph neural network (GNN), based on several technical innovations. First, we compute embeddings of all nodes by passing messages within and across the two input graphs, guided by an injective alignment between their nodes. Second, we update this alignment in a lazy fashion over multiple rounds. Within each round, we run a layerwise GNN from scratch, based on the current state of the alignment. After the completion of one round of GNN, we use the last-layer embeddings to update the alignments, and proceed to the next round. Third, IsoNet++ incorporates a novel notion of node-pair partner interaction. Traditional early interaction computes attention between a node and its potential partners in the other graph, the attention then controlling messages passed across graphs. In contrast, we consider node pairs (not single nodes) as potential partners. Existence of an edge between the nodes in one graph and non-existence in the other provide vital signals for refining the alignment. Our experiments on several datasets show that the alignments get progressively refined with successive rounds, resulting in significantly better retrieval performance than existing methods. We demonstrate that all three innovations contribute to the enhanced accuracy. Our code and datasets are publicly available at https://github.com/structlearning/isonetpp.

Problem

• Enhancing subgraph matching accuracy through iterative alignment refinement
• Developing cross-graph message passing with dynamic node-pair alignment updates
• Improving graph retrieval using novel partner interaction between node pairs

Innovation

• Early interaction GNN with cross-graph message passing
• Lazy alignment updates over multiple GNN rounds
• Node-pair partner interaction using edge existence signals