Probabilistic Models for Dynamic Hypergraphs with Content

Executive Summary

To improve the semisupervised statistical machine learning that undergirds many data-driven cybersecurity applications, we will create and develop generative models for dynamic graphs capable of extracting information from large data sets at scale using parallel estimation on high-performance computing resources. Developing large-scale machine learning models for tightly coupled information shared across time, content, and structure is applicable to an array of national security applications.