Graph Matching Networks for Learning the Similarity of Graph Structured Objects

By Y. Li et al

Published on June 10, 2019

Read the original document by opening this link in a new tab.

Table of Contents

1. Introduction
2. Related Work
3. Graph Matching Networks
4. Distance Metric Learning
5. Siamese Networks

Summary

This paper addresses the challenging problem of retrieval and matching of graph structured objects, introducing Graph Matching Networks (GMNs) for similarity learning. The authors demonstrate the effectiveness of using Graph Neural Networks (GNNs) to produce embeddings for similarity learning. GMNs utilize a cross-graph attention mechanism to compute similarity scores by associating nodes across graphs. The proposed models outperform established baselines on tasks like binary function similarity search and mesh retrieval. The paper contributes to the field by showing how GNNs can be leveraged for similarity learning and introducing the powerful GMNs for cross-graph matching.