My primary research interests lie in developing effective statistical and machine learning methods for high-dimensional “objects” with low-dimensional underlying structures. Examples of these objects include images, surfaces, networks, and time-indexed paths on non-linear manifolds, coming from neuroscience, computer vision, epidemiology, genomics, and meteorology. Most of my recent studies focus on developing novel machine learning methods to extract knowledge from large neuroimaging datasets. With advancements of in vivo brain imaging techniques, large-scale neuroimaging datasets containing more than 10k subjects can be easily accessed now. With large samples, we can gain more statistical power, a narrower margin of error, and reproducible results, but we also face modeling and computational challenges. I am dedicated to discovering efficient, elegant, and practical solutions to these challenges.