A1701
Title: Response prediction with convergence guarantees on multiple random graphs on unknown manifolds
Authors: Aranyak Acharyya - Johns Hopkins University (United States) [presenting]
Jesus Arroyo - Texas A&M University (United States)
Michael Clayton - University of Cambridge (United Kingdom)
Marta Zlatic - University of Cambridge (United Kingdom)
Youngser Park - Johns Hopkins University (United States)
Carey Priebe - Johns Hopkins University (United States)
Abstract: In real life, data in the form of multilayer networks, modeled by graphs on a common set of nodes but with edges forming randomly according to different laws at different graphs, are often encountered. In the real world, high dimensional data often admits an underlying low dimensional manifold structure. In our context, we assume that the multilayer of graphs under consideration is located on an unknown one-dimensional manifold in a higher-dimensional ambient space, with each graph corresponding to a point on the manifold. Under certain regularity assumptions, we propose a method to consistently predict graph level responses at an unlabeled graph, in a semisupervised setting, by exploiting the underlying unknown manifold structure.
