PLANETALIGN: A Comprehensive Python Library for Benchmarking Network Alignment

TL;DR

PLANETALIGN introduces a comprehensive, open source library for benchmarking network alignment across 18 built-in datasets and 14 methods spanning consistency, embedding, and optimal transport approaches. It standardizes data handling, evaluation, and experiment pipelines via unified PyTorch Geometric APIs, enabling reproducible, scalable benchmarking and method development. Empirical results show OT-based methods PARROT and JOENA often achieve state-of-the-art alignment, while embedding and consistency baselines vary by dataset, highlighting the value of hybrid strategies and robust benchmarking. The work facilitates fair comparisons, accelerates method development, and guides future NA research toward scalable and robust solutions applicable across social, biological, knowledge, and infrastructure networks.

Abstract

Network alignment (NA) aims to identify node correspondence across different networks and serves as a critical cornerstone behind various downstream multi-network learning tasks. Despite growing research in NA, there lacks a comprehensive library that facilitates the systematic development and benchmarking of NA methods. In this work, we introduce PLANETALIGN, a comprehensive Python library for network alignment that features a rich collection of built-in datasets, methods, and evaluation pipelines with easy-to-use APIs. Specifically, PLANETALIGN integrates 18 datasets and 14 NA methods with extensible APIs for easy use and development of NA methods. Our standardized evaluation pipeline encompasses a wide range of metrics, enabling a systematic assessment of the effectiveness, scalability, and robustness of NA methods. Through extensive comparative studies, we reveal practical insights into the strengths and limitations of existing NA methods. We hope that PLANETALIGN can foster a deeper understanding of the NA problem and facilitate the development and benchmarking of more effective, scalable, and robust methods in the future. The source code of PLANETALIGN is available at https://github.com/yq-leo/PlanetAlign.

Figures (6)

• Figure 1: An example of network alignment problems.
• Figure 2: An overview of built-in datasets and NA methods in PlanetAlign.PlanetAlign includes 6 categories of datasets and 3 categories of NA methods. For built-in datasets, we indicate if they consist of attributed (Attr.) or plain networks, and if they are real-world (Real) or synthetic datasets. For built-in NA methods, we indicate if they are designed for attributed or plain NA tasks, and if they are semi-supervised or unsupervised methods.
• Figure 3: Example usage of PlanetAlign for running an NA algorithm on a dataset. Users begin by instantiating the relevant dataset and algorithm objects, followed by initializing a logger for training-time logging and visualization. Model training and evaluation can then be performed through simple API calls with user-defined parameters, providing substantial flexibility in controlling the training and testing workflows.
• Figure 4: Scalability results on ER graphs.
• Figure 5: Robustness results of NA methods under different levels of edge, attribute, and supervision noises on five representative datasets.