报告标题：Edge Labels as Structural Decisions in Networks
报告时间：8月25日16:00-17:30
报告地点：理科大楼B211
报告摘要：
Networks are a fundamental model for systems ranging from social interactions and collaboration patterns to communication, infrastructure, and learning-integrated settings. Yet in many applications, knowing which edges exist is not enough: we must also assign labels that determine how the network functions. Such labels may specify when links should be activated, whether relationships are strong or weak, or how interactions should be interpreted across different contexts. This talk provides a unifying view of these problems through the lens of constrained edge labeling in networks, where edge labels are optimized under global requirements such as temporal reachability, fairness, triadic closure, cost, and cross-layer consistency.We will show how this perspective leads to algorithmic questions in both learning-integrated and social network settings. In temporal activation problems, the goal is to sparsely label edges with activation times so that different groups can fairly reach designated targets. In multilayer networks, the goal is to infer strong and weak ties that satisfy structural closure constraints while preserving a coherent interpretation of relationships across layers. Across these settings, edge labels are not independent local annotations, but structural decisions shaped by the network as a whole.
报告人简介：
Lutz Oettershagen is an Assistant Professor in the School of Computer Science and Informatics at the University of Liverpool. His research focuses on algorithmic data analysis, graph data mining, and machine learning on graphs. He received his Ph.D. in Computer Science from the University of Bonn, where his thesis on temporal graph algorithms was awarded summa cum laude. Before joining Liverpool, he was a postdoctoral researcher at KTH Royal Institute of Technology and at the LAMARR Institute/University of Bonn. His work has appeared at venues including NeurIPS, KDD, WebConf, WSDM, SDM, and ECML PKDD, and received a Best Paper Award at the International Workshop on Mining and Learning with Graphs.