TL;DR
This paper systematically explores the design space of neural graph representations for subgraph matching, revealing that novel combinations of design choices significantly improve performance and provide general principles for neural graph interaction.
Contribution
It introduces a unified framework for neural graph matching networks, thoroughly explores the design space, and uncovers new combinations that enhance performance and understanding.
Findings
Unexplored combinations in the design space yield large performance gains.
The study provides general design principles for neural graph representations.
Different interaction mechanisms between query and corpus graphs impact results.
Abstract
Subgraph matching is vital in knowledge graph (KG) question answering, molecule design, scene graph, code and circuit search, etc. Neural methods have shown promising results for subgraph matching. Our study of recent systems suggests refactoring them into a unified design space for graph matching networks. Existing methods occupy only a few isolated patches in this space, which remains largely uncharted. We undertake the first comprehensive exploration of this space, featuring such axes as attention-based vs. soft permutation-based interaction between query and corpus graphs, aligning nodes vs. edges, and the form of the final scoring network that integrates neural representations of the graphs. Our extensive experiments reveal that judicious and hitherto-unexplored combinations of choices in this space lead to large performance benefits. Beyond better performance, our study uncovers…
Peer Reviews
Decision·ICLR 2025 Poster
**S1.** The charting of the design space is convincing and well-articulated. **S2.** Empirical studies demonstrate the effectiveness of the design space in characterizing existing methods and devising superior new methods for the experiments considered. **S3.** The presentation is quite neat and clear overall.
**W1.** Most datasets considered are for small molecules despite the various applications of subgraph matching mentioned in the introduction. **W2.** There is a lack of understanding in subgraph matching performance with respect to the intrinsic challenge of the setting, e.g., highly regular graphs, different graph sizes, different feature distributions, out-of-distribution settings. Such study may be best achieved with synthetic datasets. **minor** - The use of notations $\omega$ and $\eta$ a
- The authors show a detailed and thorough literature survey on neural subgraph matching and introduce general subcomponents for subgraph matching models. - The authors conduct extensive experiments on all possible combinations within design spaces. - The authors discover a new combination of subcomponents that outperforms the existing state-of-the-art model.
- While the paper does a thorough job of exploring the design space or structured ablation studies, it does not provide a principled explanation of why certain configurations (or subcomponents) yield performance improvements. Adding a more in-depth empirical and theoretical analysis can address this issue. This can not only strengthen the contributions but also offer a more principled foundation for future research in this area. - Although the best combination shows performance benefits, its no
1. The overall problem formulation is interesting and timely considering the various design choices in subgraph matching literature. 2. Ample experimental study is performed adding credibility and rigorousity to the findings and guidelines. 3. Error bar is shown for the results, e.g. in Figure 2.
1. It seems all the 10 datasets are relatively small, e.g. up to 50 nodes. I wonder if there is a reason for not choosing much larger graphs, e.g. graphs up to 1M nodes as the target graph to be searched for (while using a small query graph of ~50 nodes). 2. Some of the paper writing can be made more clearer. E.g. “Challenging the widely held expectation that early interaction is more powerful, IsoNet’s late interaction approach outperforms GMN, even when GMN’s final score computation is made as
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
