Graph data modeling and inference

H.-T. Wai

Inferring graph structure from (behavioral) data is an important topic in data science as the relationship between nodes are often unknown. In this research, we develop novel graph signal processing model and inference methods with improved, explicit bounds on the sampling complexity. These data models stem from opinion dynamics, finance networks, and complex systems, providing the mathematical framework for information flow on a network. We test our methods on real datasets to obtain new insights about the underlying networks. In the case of opinion data, this research also focuses on applications to understand and combat the spread of fake news or adoption of new products