信息管理与商业智能系学术讲座
时 间：2023年9月26日13：30-14：30
线 上：腾讯会议室 602 886 219 / 密码 722722
线 下：管理学院 史带楼302室
题 目：A Data-Centric Perspective on Pre-Training Graph Neural Networks
主持人：复旦大学管理学院 信息管理与商业智能系 许嘉蓉青年副研究员
摘 要：
In recent years, pre-training graph neural networks has gained significant attention, with a focus on acquiring transferable knowledge for downstream tasks with unlabeled data. Despite these recent endeavors, the problem of negative transfer remains a major concern when utilizing graph pre-trained models to downstream tasks. In this study, we aim to answer two questions: (1) When to pre-train: Under what situations the “graph pre-train and fine-tune” paradigm should be adopted? (2) What to pre-train: Is a massive amount of input data really necessary, or even beneficial, for graph pre-training? Our proposed framework provides three practical applications: providing the application scope of graph pre-trained models, quantifying the feasibility of pre-training, and assistance in selecting pre-training data to enhance downstream performance.
信息管理与商业智能系
2023-9-19
-
活动讲座
-
新闻动态
-
微信头条
-
招生咨询
-
媒体视角
-
瞰见云课堂
