Graph Condensation for Real-World Graph Representation at Scale

The School of EECS is hosting the following PhD Thesis Review Seminar:

Graph Condensation for Real-World Graph Representation at Scale

Speaker: Xinyi Gao

Abstract: The rapid growth of graph data poses significant challenges in storage, transmission, and particularly the training of graph neural networks (GNNs). To address these challenges, graph condensation (GC) has emerged as an innovative solution. GC focuses on synthesising a compact yet highly representative graph, enabling GNNs trained on it to achieve performance comparable to those trained on the original large graph. The notable efficacy of GC and its broad prospects have garnered significant attention and spurred extensive research. Despite recent progress, existing GC methods remain limited in their capacity to meet the diverse and demanding requirements of real-world applications. Many are designed for idealised settings and lack the flexibility, efficiency, and robustness required for deployment in large-scale, dynamic graph environments. Motivated by this gap, this thesis takes a substantial step forward toward the principled and practical deployment of GC in real-world systems. Centred around three fundamental dimensions—generalisation, efficiency, and robustness—this thesis introduces a series of novel methods that address five core challenges constraining the current state of GC: Task-Generalisable Graph Condensation, Temporally-Invariant Graph Condensation, Training-Efficient Graph Condensation, Inference-Efficient Graph Condensation, and Structure-Robust Graph Condensation. Collectively, these contributions advance the state of graph condensation by enhancing its generalisation, efficiency, and robustness, thereby laying a solid foundation for its practical deployment in real-world systems.

Biography: Xinyi Gao is a final-year Ph.D. student at the University of Queensland, supervised by Prof. Hongzhi Yin and Dr. Rocky Chen. He received his B.S. and M.S. degrees in Information and Communications Engineering from Xi'an Jiaotong University, China. His research interests include Data-centric ML, Generative AI and Network Analysis.