In many contexts we may expect the structure of networks to be derived from some kind of abstract distance between actors. We refer to this phenomenon as homophily: like nodes connect to like. For example, people with similar beliefs may be more likely to form social relations.
We formalize this notion by positioning the nodes in a latent space representing the possible values of the homophilous attributes. Realistically, we should expect latent attributes like beliefs to change over time in some nontrivial way, and the structures of temporal networks to evolve accordingly. We introduce a model of latent space dynamics where node positions evolve according to Brownian motion. While not a realistic model for social behavior, we expect this will be a useful null model for exploring the range of possible latent space dynamics
Jasper Alt is a master's candidate in computer science at Portland State University. He is interested in networks, natural computation, and environmental prediction
System analysis, Social networks -- Mathematical models, Brownian motion processes, System theory
Alt, Jasper, "Latent Space Models for Temporal Networks" (2019). Systems Science Friday Noon Seminar Series. 73.