Learning graph edit distance by graph neural networks

• We adapt a traditional non-learnable GED algorithm to the novel paradigm of geometric deep learning. Triplet network for learning graph distances by means neural networks. Learning distance in domain without an embedding stage. Graph-based keyword spotting application with state-of-the-art performance. The emergence as framework deal graph-based representations has faded away approaches favor completely new methodologies. In this paper, we propose able combine advances on metric approximations edit distance. Hence, efficient based field Our method employs message passing capture structure, and thus, leveraging information its use computation. performance proposed is validated two different scenarios. On one hand, retrieval handwritten words i.e. spotting, showing superior when compared (approximate) benchmarks. other demonstrating competitive results similarity current recent benchmark dataset.