Working on the development of sparse statistical modelling techniques.

Xinle completed his undergraduate degree in Mathematics at the University of British Columbia and his Master’s degree at the George Washington University in Statistics.

Research project title:
Development of Sparse Statistical Modelling for Neurological Applications

Supervisor(s): 

Sandipan Roy, Matthew Nunes

Project description:
The aim of many neurological studies, for example using fMRI data, is to derive the areas of the brain and connections between them which are the principal contributors to brain activity. Current techniques often focus on latent variable models to quantify the main features of process dynamics, but can often produce negative weightings for components, contrary to scientific belief and resulting in models which are difficult to interpret. This project aims to develop new models for functional activity which don’t enforce restrictive model assumptions, with associated estimation theory and fitting based on constrained optimization to ensure clinically-relevant parameter values. In addition, we will develop inference methods in which structure can be included as a graph prior for sparse graphical models. Subsequent research will then focus on algorithms for detecting changes in such settings. With the potential new models, it serves to both improve interpretability and predictive performance in neurological studies and allow us to investigate a better understanding of functional connectivity which occurs in neurological studies.