Core AI Algorithm Developments

We will focus on several theory and algorithmic questions to develop core AI capabilities that will have significant translational impact. These directions include developing robust and interpretable machine learning (ML) algorithms, multi-modal fusion, domain knowledge accommodation, semi-/unsupervised/reinforcement learning.

Physics Informed Machine Learning

Application of AI in many science and engineering problems requires incorporation of physics (first principles) of the problems into AI models to guarantee physically meaningful model outcomes as well as reduce training data requirement. Our team members have made pioneering contributions in this emerging field of study.

Robust and Interpretable Machine Learning

We explore robustness, interpretability and learning from heterogeneous data types such as spatial, temporal and spatiotemporal data. In addition, we also explore natural language processing methods to analyze language-based biological data to construct knowledge graphs that assist in generation of testable hypotheses for genetics and agricultural applications.

Neural-symbolic AI

We explore robustness, interpretability and learning from heterogeneous data types such as spatial, temporal and spatiotemporal data. In addition, we also explore natural language processing methods to analyze language-based biological data to construct knowledge graphs that assist in generation of testable hypotheses for genetics and agricultural applications.