Research
I work on various kinds of inverse problems related to scientific machine leanring:
- Learing Self Organization from Observation
- Physics Informed Machine Learning for Solving and Learning PDEs
Learning Collective Behaviors
We have developed series of variational inverse problem based learning methods to explain collective behaviors (also known as self organization), such as clustering, flocking, swarming, and synchronization, from observation data. A summary presentation of the research can be found here, a follow up presentation can be found here, and a webinar on this topic can be found here.
Physics Informed Machine Learning
We have developed a series of training schemes for solving PDE driven dynamics using Physics Informed Neural Networks as well as Physics Informed Gaussian Processes. An overview of the research can be found here, and a webinar on this topic can be found here.