Our main area of research is in statistical machine learning. The core problem here is to infer conclusions from observations or data. The caveat is to do so reliably with limited computation and limited data.

We are especially interested in the frameworks of probabilistic graphical models, and high-dimensional statistical models. Graphical Models compactly represent distributions over a large number of variables using undirected graphs which encode conditional independence assumptions among the variables. High-dimensional Statisical Models are increasingly indispensable in modern settings where the dimensionality of the data is high when compared with the number of observations: the hope is to leverage some low-dimensional structure such as sparsity, low-rank, manifold structure, etc. These frameworks are used in domains across engineering such as search engines, medical diagnosis, image processing, speech recognition, as well as various empirical and natural sciences.

     [Expand to show all 40][Minimize]