Strong Transitivity Relations and Graph Neural Networks

TL;DR

This paper introduces TransGNN, a novel graph neural network that captures both local and global similarities through transitivity relations, enhancing node classification performance.

Contribution

It extends similarity concepts in GNNs to include transitivity relations, enabling the modeling of global similarities across the entire graph.

Findings

TransGNN outperforms existing GNN models on multiple datasets.

Global similarity modeling improves node classification accuracy.

Distinguishing strong and weak transitivity relations is effective.

Abstract

Local neighborhoods play a crucial role in embedding generation in graph-based learning. It is commonly believed that nodes ought to have embeddings that resemble those of their neighbors. In this research, we try to carefully expand the concept of similarity from nearby neighborhoods to the entire graph. We provide an extension of similarity that is based on transitivity relations, which enables Graph Neural Networks (GNNs) to capture both global similarities and local similarities over the whole graph. We introduce Transitivity Graph Neural Network (TransGNN), which more than local node similarities, takes into account global similarities by distinguishing strong transitivity relations from weak ones and exploiting them. We evaluate our model over several real-world datasets and showed that it considerably improves the performance of several well-known GNN models, for tasks such as node classification.