Scientific Machine Learning

Our research spans to:

• Data-driven simulation methods, where raw data are incorporated into physics-based numerical simulations;
• Physics-informed machine learning, where physical constraints are integrated into data-driven machine learning techniques;
• Applications in challenging engineering design tasks, such as optimization, uncertainty quantification, and parameter identification.

• Data-driven, material-model-free electromagnetic field solvers
• Numerical simulation with physics-informed neural networks
• Bayesian inference methods for parameter identification in biological and electrical engineering systems
• Tensor decompositions for solving high-dimensional multi-linear systems
• Uncertainty quantification for models with high-dimensional inputs and outputs
• Surrogate-based design optimization