Abstract:Graph alignment considers the optimal node correspondence across networks. To advance unsupervised graph alignment algorithms on plain graphs, we propose Degree Matrix Comparison (DMC). Through extensive experiments and mathematical motivations, we demonstrate the potential of this method. Remarkably, DMC achieves up to 99% correct node alignment for 90%-overlap graphs and 100% accuracy for isomorphic graphs. Additionally, we propose a reduced version of DMC (Greedy DMC) that provides a solution to the graph alignment problem with lower time complexity. DMC could significantly impact graph alignment, offering a reliable solution for the task.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the Graph Alignment problem. Specifically, it is to perform node correspondence for simple un - attributed graphs in an unsupervised manner. The graph alignment task has important applications in multiple fields, such as social networks, biology, and network security. Some existing methods rely on node attributes or semi - supervised information, but in cases where users disguise their identities or lack node attributes, the effectiveness of these methods will be greatly reduced. For this reason, the authors propose the **Degree Matrix Comparison (DMC)** method, aiming to achieve node alignment by comparing the degree matrices of two graphs. In addition, they also propose a simplified version of DMC, called **Greedy DMC**, to reduce the time complexity. ### Main Contributions 1. **Proposing the DMC method**: By comparing the degree matrices of two graphs, unsupervised graph alignment is achieved. 2. **Experimental verification**: The effectiveness of the DMC method has been verified through extensive experiments, especially achieving a correct alignment rate of up to 99% on highly overlapping graphs and isomorphic graphs. 3. **Greedy DMC**: A simplified version of DMC is provided, which can significantly reduce the time complexity while ensuring a certain accuracy. ### Theoretical Motivations - **Invariance of the degree matrix**: The degree matrix remains invariant under any node labels. - **Compact representation**: Compared with the adjacency matrix, the degree matrix (N×m) can encode information more compactly. - **Combination of local and global perspectives**: Combining local (node degree) and global (node assignment) perspectives helps to better capture the structural characteristics of the graph. ### Experimental Results The experimental results show that DMC performs well on a variety of datasets, especially on protein - protein interaction networks (PPI) and biological networks. The accuracy of DMC on the classic PPI network is higher than that of other baseline methods, and it performs best on heterogeneous graphs. ### Future Work Future research can further explore the performance of Greedy DMC in different situations and try to find a better balance point to reduce the time complexity while ensuring accuracy. In conclusion, this paper provides a new solution to the unsupervised graph alignment problem by proposing the DMC method and shows its potential in various application scenarios.

Degree Matrix Comparison for Graph Alignment

An Entity Alignment Method Based on Graph Attention Network with Pre-classification

Deep graph alignment network

On the Feasible Region of Efficient Algorithms for Attributed Graph Alignment

Comparison of Molecule Graph Representation with Similarity Consistency.

Efficient Algorithms for Attributed Graph Alignment with Vanishing Edge Correlation

Network alignment based on multiple hypernetwork attributes

Cross-graph Embedding with Trainable Proximity for Graph Alignment

Graph Alignment Neural Network Model With Graph to Sequence Learning

GABoost: Graph Alignment Boosting via Local Optimum Escape

Comparative Evaluation of Bipartite, Node-Link, and Matrix-Based Network Representations

Graph matching for knowledge graph alignment using edge-coloring propagation

Attributed Network Alignment: Problem Definitions and Fast Solutions

DAHGN: Degree-Aware Heterogeneous Graph Neural Network

The graph alignment problem: fundamental limits and efficient algorithms

Pairwise Alignment Improves Graph Domain Adaptation

Spectral Alignment of Graphs

Efficient Random Graph Matching Via Degree Profiles

Triangular Alignment (TAME): A Tensor-based Approach for Higher-order Network Alignment

GANN: Graph Alignment Neural Network for semi-supervised learning

Graduated Consistency-Regularized Optimization For Multi-Graph Matching