![[2602.22645] MUG: Meta-path-aware Universal Heterogeneous Graph Pre-Training](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.22645] MUG: Meta-path-aware Universal Heterogeneous Graph Pre-Training
Summary
The paper presents MUG, a novel approach for universal heterogeneous graph pre-training, addressing challenges in encoding diverse graph structures and semantics for improved representation learning.
Why It Matters
As graph representation learning becomes increasingly critical in machine learning, MUG offers a solution for effectively handling heterogeneous graphs, which are prevalent in real-world applications. This advancement could enhance the performance of various downstream tasks, making it relevant for researchers and practitioners in the field.
Key Takeaways
- MUG introduces a unified representation space for heterogeneous graphs.
- The approach addresses challenges in encoding diverse meta-paths across datasets.
- Extensive experiments show MUG's effectiveness on real datasets.
- MUG's design improves generalization across different graph types.
- The method reduces dataset-specific biases in graph representation.
Computer Science > Machine Learning arXiv:2602.22645 (cs) [Submitted on 26 Feb 2026] Title:MUG: Meta-path-aware Universal Heterogeneous Graph Pre-Training Authors:Lianze Shan, Jitao Zhao, Dongxiao He, Yongqi Huang, Zhiyong Feng, Weixiong Zhang View a PDF of the paper titled MUG: Meta-path-aware Universal Heterogeneous Graph Pre-Training, by Lianze Shan and 5 other authors View PDF HTML (experimental) Abstract:Universal graph pre-training has emerged as a key paradigm in graph representation learning, offering a promising way to train encoders to learn transferable representations from unlabeled graphs and to effectively generalize across a wide range of downstream tasks. However, recent explorations in universal graph pre-training primarily focus on homogeneous graphs and it remains unexplored for heterogeneous graphs, which exhibit greater structural and semantic complexity. This heterogeneity makes it highly challenging to train a universal encoder for diverse heterogeneous graphs: (i) the diverse types with dataset-specific semantics hinder the construction of a unified representation space; (ii) the number and semantics of meta-paths vary across datasets, making encoding and aggregation patterns learned from one dataset difficult to apply to others. To address these challenges, we propose a novel Meta-path-aware Universal heterogeneous Graph pre-training (MUG) approach. Specifically, for challenge (i), MUG introduces a input unification module that integrates informati...
