Learning and Inference on Graph-structured Data
Wei Ye, the University of California, Santa Barbara
2020-05-13 14:00:00 ~ 2020-05-13 15:00:00
Graph-structured data is ubiquitous in the real-world. Because graph is a natural way to represent the relationships between objects, it has attracted a lot of attention. However, graph-structured data have no spatial or temporal order inside. Standard deep learning architectures such as CNNs and RNNs cannot directly work on them.
In the first part of this talk, I will present shallow and deep models on graph-structured data for the tasks of local clustering, semi-supervised node classification, and graph classification. In the second part of this talk, I will present an inference model on the appraisal network among a hybrid team of humans and AI agents. The developed inference model interprets human-AI decision making on truth statements well. In the last part of my talk, I will discuss my future research directions including relational and causal inference by deep neural networks and the theory of human/AI’s mind in designing hybrid human-AI teams.
Wei Ye is currently a postdoctoral researcher with the DYNAMO lab at the University of California, Santa Barbara. Before joining the DYNAMO lab, he worked as a researcher in the Department of AI Platform, Tencent, China. He obtained his PhD degree in Computer Science from Institute for Informatics, Ludwig-Maximilians University of Munich, Germany, in 2018. His research interests include graph machine learning and their applications, causal inference and reasoning, and dynamic networks.