UB Faculty Member Leads DARPA AI Project

     The principal goal of the project is to develop novel hybrid methods that combine Multi-physics equation based models with data-driven machine learning models (such as Deep Neural Networks (DNNs)) to enable predictive modeling of complex systems in the presence of imperfect models and sparse and noisy data. Since the physical model captures in part the system  behavior, the need for a complex data-driven model is reduced (i.e., we have fewer layers and fewer parameters). In addition, the reduced number of parameters makes the training algorithm more amenable to sparse data.