AutoGL: A Library for Automated Graph Learning

TL;DR

This work proposes a three-layer architecture, consisting of backends to interface with devices, a complete automated graph learning pipeline, and supported graph applications, and presents AutoGL-light, a lightweight version of AutoGL to facilitate customizing pipelines and enriching applications.

Abstract

Recent years have witnessed an upsurge in research interests and applications of machine learning on graphs. However, manually designing the optimal machine learning algorithms for different graph datasets and tasks is inflexible, labor-intensive, and requires expert knowledge, limiting its adaptivity and applicability. Automated machine learning (AutoML) on graphs, aiming to automatically design the optimal machine learning algorithm for a given graph dataset and task, has received considerable attention. However, none of the existing libraries can fully support AutoML on graphs. To fill this gap, we present Automated Graph Learning (AutoGL), the first dedicated library for automated machine learning on graphs. AutoGL is open-source, easy to use, and flexible to be extended. Specifically, we propose a three-layer architecture, consisting of backends to interface with devices, a complete automated graph learning pipeline, and supported graph applications. The automated machine learning pipeline further contains five functional modules: auto feature engineering, neural architecture search, hyper-parameter optimization, model training, and auto ensemble, covering the majority of existing AutoML methods on graphs. For each module, we provide numerous state-of-the-art methods and flexible base classes and APIs, which allow easy usage and customization. We further provide experimental results to showcase the usage of our AutoGL library. We also present AutoGL-light, a lightweight version of AutoGL to facilitate customizing pipelines and enriching applications, as well as benchmarks for graph neural architecture search. The codes of AutoGL are publicly available at https://github.com/THUMNLab/AutoGL.

Figures (2)

• Figure 1: An overall framework of AutoGL. The architecture is divided into three layers. At the bottom layer, AutoGL uses the existing graph learning libraries PyTorch Geometric pyg and Deep Graph Learning dgl as backends to interact with the hardware devices. At the middle layer, we design a comprehensive automated graph learning solution. First, we use AutoGL Dataset class to manage diverse graph data, including homogeneous and heterogeneous graphs. Then, we utilize AutoGL Solver, a high-level API to control the workflow of automated graph learning, covering five main functional blocks: Auto Feature Engineering, Neural Architecture Search, Hyper-Parameter Optimization, Model Training, and Auto Ensemble. Our design covers the mainstream automated graph learning methods. At the top layer, we support various graph tasks, including node classification, link prediction, graph classification, self-supervised graph learning, robust graph learning, heterogeneous node classification, etc.
• Figure 2: An overall framework of AutoGL-light. Compared to AutoGL, AutoGL-light is a more lightweight version with decoupled designs and focuses on two core functionalities of automated graph machine learning: graph HPO and graph NAS. Through modularized functions, AutoGL-light can more flexibly support customized pipelines. We also refactor the code to be more user-friendly to PyG users.