I work on the theory and application of machine learning, especially for large-scale spatiotemporal data. I am generally interested in optimization, deep learning, and spatiotemporal reasoning. I am particularly excited about the interplay between physics and machine learning. My work has been applied to learning dynamical systems, sports analytics, intelligent transportation, and climate informatics. For more details, see my curriculum vitae. For prospective students, please read this before emailing me.
Our physical world is intrinsically spatiotemporal. We design deep learning models that bridge the expressiveness of neural networks and the rich spatiotemporal structures from the data, addressing fundamental challenges of high-dimensionality, high-order correlation, non-linear dynamics and multi-resolution dependencies.
A graphical model can conveniently encode high-level structures. The nodes of the graph typically represent the problem components, and the edges capture their spatiotemporal interactions. We aim to study a novel deep learning framework with a primary focus on combining graphical models with black-box deep neural networks.
Providing interpretable insights for domain scientists is one of the core missions in data-driven decision making. We leverage the efficiency and simplicity of tensor factor models to uncover the hidden spatiotemporal patterns from the data.