Project Grant 2204936

This three-year Project Grant from the National Science Foundation's Mathematical and Physical Sciences program, totaling $300,000, will support the development of temporal network embedding methods to map time-varying network data to trajectories in a latent space. Specifically, the awardee will establish a family of temporal network embedding techniques that represent the network structure at a given time point succinctly as a point on the trajectory. By embedding entire networks rather than individual nodes, the proposed methods aim to capture gross properties of temporal data and enable applications like visualization, anomaly detection, and discovery of periodic patterns. The awardee will build mathematical foundations for the techniques, applying them to social, financial, bibliographic, neuroimaging, and climate data. Outcomes may encourage further algorithm development in the network embedding field. Products include mathematical models, software implementations, and empirical analyses delivering insights from temporal network data in domains important to the NSF's mission to advance the sciences.