报告题目：Anomaly Detection Foundation Models for Structured Data
报告摘要： Anomaly detection foundation models offer a promising direction for building generalist detectors across heterogeneous structured datasets. This talk presents recent advances in anomaly detection for graph and tabular data, including ARC (NeurIPS’24), which uses in-context learning to detect anomalies on unseen graphs; REFI-GAD (ICML’26), which captures transferable anomaly patterns through relational fingerprints; and OFA-TAD (ICML’26), which combines transferable neighbor-distance patterns with multi-view anomaly evidence. The talk concludes with a discussion of future directions for anomaly detection foundation models.
报告人简介： 刘奕鑫，男，澳大利亚基金委优秀青年基金（ARC DECRA Fellow）、Google博士奖学金（Google PhD Fellow）获得者。现任澳大利亚格里菲斯大学信息与通信技术学院（School of ICT, Griffith University）助理教授。刘博士于2025和2026年入选斯坦福-艾斯维尔全球前2%顶尖科学家榜单（the World’s Top 2% Scientists）。刘博士在NeurIPS、ICML、ICLR、KDD、TPAMI、TKDE、TNNLS等刊物和会议上发表高水平论文60余篇，同时担任TNNLS、TKDE、TIP等领域期刊审稿人，任WWW、IJCAI、AAAI、CIKM、ICDM等顶级会议程序委员会委员，谷歌学术引用6000+。主要研究方向为数据挖掘、机器学习，侧重于图神经网络、异常检测、AI Agent等方向。
地点：吉林大学正新楼三楼智慧教室
时间：2026年11月11日（星期三）上午9:30-11:00
