Novel Approaches for Mining Large and Complex Networks
This CAREER project includes an integrated research, education, and outreach program that focuses on the development of novel methods for mining large, complex networks. Networks (graphs) are ubiquitous in real-world applications. Although successful, the methodology development for network analytics is still in its early stage. This project addresses fundamental questions essential to the advancement of large and complex network analytics. These challenges are driven by real-world applications in social, biological, and medical domains. The research plan is complemented by a comprehensive education and outreach plan focused on three elements: (1) the development of new interdisciplinary courses; (2) direct undergraduate involvement in the research projects; and (3) outreach activities including the STEM program targeting K-12 schools. Underrepresented students will be encouraged to participate in this project.
The research goal of the project is to significantly extend the reliability and efficiency of large network analysis. The project has three research aims. (1) Develop novel memory-based random walk proximity measures that can effectively capture the similarity between nodes. Proximity measure is of fundamental importance for many advanced network analysis tasks. A rigorous theoretical foundation will be provided for the developed measures. (2) Study the dual-network model and its applications. The dual-network model has a wide range real-world applications. The specific problem of cross-network node set query will be investigated in the project. Both numerical and algorithmic approaches will be explored. (3) Design robust and flexible multi-network algorithms for clustering and ranking. The focus will be on a novel multi-network model, a network of networks, which allows us to integrate domain similarities to improve the performance of the algorithms.