NAS-Bench-Graph: Benchmarking Graph Neural Architecture Search

TL;DR

This work constructs a unified, expressive yet compact search space, covering 26,206 unique graph neural network (GNN) architectures and proposes a principled evaluation protocol, which enables fair, fully reproducible, and efficient comparisons for GraphNAS.

Abstract

Graph neural architecture search (GraphNAS) has recently aroused considerable attention in both academia and industry. However, two key challenges seriously hinder the further research of GraphNAS. First, since there is no consensus for the experimental setting, the empirical results in different research papers are often not comparable and even not reproducible, leading to unfair comparisons. Secondly, GraphNAS often needs extensive computations, which makes it highly inefficient and inaccessible to researchers without access to large-scale computation. To solve these challenges, we propose NAS-Bench-Graph, a tailored benchmark that supports unified, reproducible, and efficient evaluations for GraphNAS. Specifically, we construct a unified, expressive yet compact search space, covering 26,206 unique graph neural network (GNN) architectures and propose a principled evaluation protocol. To avoid unnecessary repetitive training, we have trained and evaluated all of these architectures on nine representative graph datasets, recording detailed metrics including train, validation, and test performance in each epoch, the latency, the number of parameters, etc. Based on our proposed benchmark, the performance of GNN architectures can be directly obtained by a look-up table without any further computation, which enables fair, fully reproducible, and efficient comparisons. To demonstrate its usage, we make in-depth analyses of our proposed NAS-Bench-Graph, revealing several interesting findings for GraphNAS. We also showcase how the benchmark can be easily compatible with GraphNAS open libraries such as AutoGL and NNI. To the best of our knowledge, our work is the first benchmark for graph neural architecture search.

Figures (12)

• Figure 1: An illustration of the 9 different choices of our macro search space. Each node indicates a representation of vertices and each edge indicates an operation. We omit the output node for clarity.
• Figure 2: The distribution of accuracy, latency, and the numbers of parameters of all architectures. The Pareto-optimal architectures w.r.t. accuracy and latency are marked with red crosses.
• Figure 3: The frequency of the macro space and operation choices in the top 5% architectures of different datasets. Please refer to Figure \ref{['fig:macrospace']} for the macro space choices.
• Figure 4: The architecture performance correlation across different datasets using three metrics.
• Figure 5: The performance difference between architectures with different number of mutations. The red lines indicate the average performance difference between two random architectures.