Graph Similarity

Graph similarity research focuses on developing methods to quantify the resemblance between graph-structured data, enabling comparisons and analyses across diverse applications. Current research emphasizes efficient algorithms for computing graph similarity metrics like graph edit distance, often leveraging graph neural networks (GNNs) and techniques such as contrastive learning and belief propagation to learn effective graph representations. These advancements are crucial for improving performance in various fields, including node classification, clustering, change-point detection in dynamic networks, and even software analysis through binary diffing. The development of accurate and scalable graph similarity methods is driving progress in numerous domains requiring analysis of relational data.