Are More Layers Beneficial to Graph Transformers?

TL;DR

This paper investigates the depth limitations of graph transformers, identifies the bottleneck caused by global attention, and proposes DeepGraph, a model that incorporates substructure tokens and local attention to enable deeper, more expressive graph transformers with state-of-the-art results.

Contribution

The paper introduces DeepGraph, a novel graph transformer architecture that uses substructure tokens and local attention to overcome depth limitations and improve expressiveness.

Findings

DeepGraph enables deeper graph transformer models.

The proposed method achieves state-of-the-art performance on various benchmarks.

DeepGraph effectively focuses on critical substructures within graphs.

Abstract

Despite that going deep has proven successful in many neural architectures, the existing graph transformers are relatively shallow. In this work, we explore whether more layers are beneficial to graph transformers, and find that current graph transformers suffer from the bottleneck of improving performance by increasing depth. Our further analysis reveals the reason is that deep graph transformers are limited by the vanishing capacity of global attention, restricting the graph transformer from focusing on the critical substructure and obtaining expressive features. To this end, we propose a novel graph transformer model named DeepGraph that explicitly employs substructure tokens in the encoded representation, and applies local attention on related nodes to obtain substructure based attention encoding. Our model enhances the ability of the global attention to focus on substructures and promotes the expressiveness of the representations, addressing the limitation of self-attention as the graph transformer deepens. Experiments show that our method unblocks the depth limitation of graph transformers and results in state-of-the-art performance across various graph benchmarks with deeper models.