Graph Theory Problems

Using Deep Learning for graph theory and subgraph matching

Graph theory is a branch of mathematics that studies graphs, which are mathematical structures used to model pairwise relations between objects. Currently, existing classical algorithms can solve these problems effectively. However, these algorithms are not scalable and cannot be applied to large-scaled graphs.

Research Focus

I explore how to use Deep Learning to solve graph theory problems with targeted large-scaled graphs and applications in drug discovery, social network analysis, etc.

Master’s Thesis

My Master’s thesis focuses on enhancing subgraph isomorphism prediction models to improve scalability and interpretability, particularly for applications in drug design. I introduce xNeuSM, an Explainable Neural Subgraph Matching framework, which leverages Graph Learnable Multi-hop Attention Networks (GLeMA) to dynamically learn attention decay parameters across multiple hops.

Key Results

  • Higher accuracy than approximate baselines
  • Query times at least 7x faster than exact algorithms
  • Practical for large-scale molecular analysis in drug discovery

Resources

  • Master’s thesis PDF: Link
  • Master’s thesis presentation: Link
  • Master’s thesis code: Link

Publications

This work has been published in two key papers, including (Nguyen et al., 2023) and (Nguyen et al., 2024).

References

2024

  1. Explainable Neural Subgraph Matching With Learnable Multi-Hop Attention
    Duc Q. Nguyen, Thanh Toan Nguyen, Jun Jo, and 3 more authors
    IEEE Access, Oct 2024

2023

  1. 10X Faster Subgraph Matching: Dual Matching Networks with Interleaved Diffusion Attention
    Thanh Toan Nguyen, Duc Q. Nguyen, Zhao Ren, and 3 more authors
    In 2023 International Joint Conference on Neural Networks (IJCNN), Jun 2023
    ISSN: 2161-4407