This $300,000 Project Grant from the National Science Foundation Division of Mathematical Sciences under the Mathematical and Physical Sciences program (CFDA 47.049) will fund research into spectral methods for single and multiple graph inference networks. The grantee, North Carolina State University, will develop efficient parameter estimation methods for latent position graphs and valid two-sample testing procedures for comparing latent position graphs while ignoring irrelevant features. The university will also conduct perturbation analysis of randomized singular value decomposition for dimension reduction of large, noisy graphs. Results will be applied to general matrix-valued data and software packages developed for network analysis. The research aims to advance network science methods for classifying networks, dimension reduction, and comparing graph similarities. Open-source software and graduate student training will disseminate findings to practitioners and next generations of data scientists.
Generated 1/7/24, 2:29 AM