The School of EECS is hosting the following HDR Progress Review 1 Confirmation Seminar:
Empowering Graph Representation Learning at Test-time
Speaker: Yan Jiang
Abstract: Test-time distribution shifts between training and testing graphs present a major challenge to the effectiveness of pre-trained models in graph representation learning. Recent data-centric approaches have explored test-time graph transformations to alleviate this issue and improve the performance of Graph Neural Networks (GNNs). However, these methods often neglect key structural properties such as homophily, limiting their effectiveness. To address this, we propose HoST, a novel data-centric test-time graph transformation method that exploits homophily-related patterns to guide graph structural adaptation at test-time and boost GNN performance.
Furthermore, Graph Foundation Models (GFMs) have demonstrated strong generalisation across tasks and domains. Recent developments in graph prompt tuning have further improved GFMs by adapting auxiliary prompts in few-shot settings. Nonetheless, existing approaches rely solely on labelled training data, ignoring unlabelled test nodes and thus failing to address train–test distribution shifts. To overcome this, we propose GFMate, a complementary learning strategy that incorporates both labelled and unlabelled target data to adapt prompts in a pre-training-agnostic manner, mitigating distribution shifts during GFM downstream adaptation.
Together, these two approaches contribute to enhancing the graph representation learning by developing pioneering test-time methods and advancing the field toward a more effective and robust graph learning system.
Bio: Yan Jiang is a PhD student in the School of Electrical Engineering and Computer Science at the University of Queensland. He earned both his Bachelor of Computer Science and Master of Data Science degrees from UQ. He focuses on the research topic of graph representation learning at test-time, under the supervision of Professor Helen Huang, Associate Professor Guangdong Bai, and Dr. Ruihong Qiu.
About Data Science Seminar
This seminar series is hosted by EECS Data Science.
Venue
Room: 78 - 632
Zoom: https://uqz.zoom.us/j/88947620086
Zoom: https://uqz.zoom.us/j/88947620086