Scalable probabilistic inference for high dimensional structured variables

Modern high-throughput data collection technologies have transformed the scope and scale of data analysis in the biological sciences, providing an unprecedented opportunity to investigate important scientific questions and address life threatening diseases with individually tailored treatments. Meaningful analysis of such high dimensional datasets must incorporate knowledge gleaned from expertise, experimental evidence and statistical considerations. In particular, restricting the degrees of freedom via sparsity or low rank structure has become an important design paradigm, enabling the recovery of parsimonious and interpretable results, and improving storage and prediction efficiency for high dimensional problems. I will describe a novel family of methods for scalable probabilistic inference with structured variables. The approach is based on analyzing the information projection of probability distributions to structured sets. When applied to variable selection, the information projection corresponds to a submodular optimization. As a result, greedy forward selection is efficient with strong optimization guarantees on the quality of the solution. I will then outline some recent successes of the proposed family of methods in neuroimaging data analysis including voxel selection for decoding task fMRI and sparse PCA for exploratory analysis of resting state fMRI.
Sanmi (Oluwasanmi) Koyejo is an engineering research associate in the Poldrack Lab at Stanford University. His research involves the development and analysis of principled methods for elucidating patterns in neuroimaging, genetics and other large-scale biological data. He completed his Ph.D in Electrical Engineering at the University of Texas at Austin under the supervision of Joydeep Ghosh and was a postdoc with Russell Poldrack and Pradeep Ravikumar. Sanmi has been the recipient of several awards including the outstanding NCE/ECE student award, a best student paper award from the conference on uncertainty in artificial intelligence (UAI) and a trainee award from the Organization for Human Brain Mapping (OHBM).