摘要
Recently, ``pre-training and fine-tuning'' has been adopted as a standard workflow for many graph tasks since it can take general graph knowledge to relieve the lack of graph annotations from each application. However, graph tasks with node level, edge level, and graph level are far diversified, making the pre-training pretext often incompatible with these multiple tasks. This gap may even cause a ``negative transfer'' to the specific application, leading to poor results. Inspired by the prompt learning in natural language processing (NLP), which has presented significant effectiveness in leveraging prior knowledge for various NLP tasks, we study the prompting topic for graphs with the motivation of filling the gap between pre-trained models and various graph tasks.
In this talk, we introduce meta-learning to efficiently learn a better initialization for the multi-task prompt of graphs so that our prompting framework can be more reliable and general for different tasks. We conduct extensive experiments, results from which demonstrate the superiority of our method.
分享嘉宾
Dr. Xiangguo Sun is now working as a postdoctoral research fellow at the Chinese University of Hong Kong. He was recognized as the "Social Computing Rising Star" in 2023 from CAAI. He studied at Zhejiang Lab as a visiting researcher in 2022. In the same year, he received his Ph.D. from Southeast University and won the Southeast University Distinguished Ph.D. Dissertation Award. During his Ph.D. study, he worked as a research intern at Microsoft Research Asia from Sep 2021 to Jan 2022, and won the ``Award of Excellence'' from MSRA in the Stars of Tomorrow Program. He also studied as a joint Ph.D. student at The University of Queensland from Sep 2019 to Sep 2021, Australia. His research interest is social computing and network learning. He was the winner of the Best Research Paper Award at KDD'23. He has published 11 CORE A*, 9 CCF A, and 13 SCI (including 6 IEEE Trans),.
