Automatic Graph Topology-Aware Transformer

TL;DR

An evolutionary graph Transformer architecture search (EGTAS) framework to automate the construction of strong graph Transformers, and a surrogate model based on generic architectural coding is proposed to directly predict the performance of graph Transformers, substantially reducing the evaluation cost of evolutionary search.

Abstract

Existing efforts are dedicated to designing many topologies and graph-aware strategies for the graph Transformer, which greatly improve the model's representation capabilities. However, manually determining the suitable Transformer architecture for a specific graph dataset or task requires extensive expert knowledge and laborious trials. This paper proposes an evolutionary graph Transformer architecture search framework (EGTAS) to automate the construction of strong graph Transformers. We build a comprehensive graph Transformer search space with the micro-level and macro-level designs. EGTAS evolves graph Transformer topologies at the macro level and graph-aware strategies at the micro level. Furthermore, a surrogate model based on generic architectural coding is proposed to directly predict the performance of graph Transformers, substantially reducing the evaluation cost of evolutionary search. We demonstrate the efficacy of EGTAS across a range of graph-level and node-level tasks, encompassing both small-scale and large-scale graph datasets. Experimental results and ablation studies show that EGTAS can construct high-performance architectures that rival state-of-the-art manual and automated baselines.

Figures (8)

• Figure 1: The overview of EGTAS consists of three components: problem definition, search space, and search strategy. According to the problem definition, the search space is carefully designed, encompassing both macro and micro levels. Then, a surrogate-assisted evolutionary search is presented to explore the search space to automate the construction of optimal graph Transformer architectures. Surrogate models are employed to predict the performance of architectures, thus reducing the computational cost of evaluations during the evolutionary search.
• Figure 2: A general architecture of the graph Transformer.
• Figure 3: An illustrative example of the encoding strategy.
• Figure 4: Evaluation process of an architecture $\alpha$.
• Figure 5: An illustrative example of a reproduction operator involving crossover and mutation.