Alexander Jung

My current research is focused on machine learning from massive network-structured data (``big data over networks''). Big data over networks refers to large collections of (billions of) local datasets that are related by an intrinsic network structure. A timely application domain that generates big data over networks is the management of pandemics. Humans generate local datasets via their social media activities and their wearables including smartphones that measure biophysical parameters. These local datasets are related via different network structures that reflect physical, social orbiological proximity. We study methods that capitalize jointly on the information in local datasets and their network structure.

To jointly leverage the information conveyed by high-dimensional data points and their network structure, we have proposed networked exponential families (preprint here ). For the accurate learning of such networked exponential families, we borrow statistical strength, via the intrinsic network structure, across the dataset.

A powerful algorithmic toolbox for designing learning algorithms is provided by convex optimization methods. Modern convex optimization methods are appealing for big data applications as they can be implemented via highly scalable message passing protocols.

A recent key result obtained from our research are precise conditions on the network structure and local datasets such that accurate predictions can be obtained by efficient methods:
A. Jung and N. Tran. Localized Linear Regression in Networked Data. IEEE Sig. Proc. Letters , 2019. Preprint