Domain-Aware Statistical Learning
Interpretable models, actionable insights, high-stakes applications
Research Objectives
To create a domain-aware statistical learning framework that enables the intrusive integration of governing physics and fundamental engineering domain knowledge into data-driven models.
To innovate new methodologies for inherently interpretable domain-aware modeling, data-driven discovery of governing physics, adaptive reduced-order modeling, and optimal dynamic sampling for non-stationary engineering/scientific processes.
The modern fields of engineering present an unparalleled data-rich and domain-knowledge-intensive environment: Governing physics and engineering domain knowledge impose critical constraints on how data can be modeled and how models should be interpreted.
The continuing penetration (of data sciences) into the modern fields of engineering can be significantly accelerated by harnessing the convergence of engineering domain knowledge into data-driven methodologies.
Inter-disciplinary Applications
Modeling of wildfires and solar energy using remote sensing data
Statistical modeling of airborne collision between aircraft and UAVs
inverse modeling: emission source detection using sensor data streams
Urban environmental processes (air pollution)