To main content

Physics Informed Machine Learning

Physics-Informed Machine Learning (PIML) is an emerging field that integrates physical knowledge directly into Machine Learning ML models. By embedding physical constraints typically described by differential equations, PIML ensures predictions do not rely only on data but are also consistent with known physical principles. This synergy between physics and ML enhances accuracy, interpretability, and generalization, making PIML particularly valuable in scenarios where data is scarce, noisy, or expensive to obtain.

Contact person

Our expertise lies in developing PIML models and methodologies to address complex challenges across science and engineering. Collaborating closely with domain experts, we incorporate physical knowledge into our frameworks. When physical knowledge is incomplete, we design hybrid models that combine physics-based constraints with data-driven learning. This allows our models to respect known physical principles while leveraging data to address uncertainties and missing information, resulting in robust and interpretable solutions.

PIML is especially impactful in engineering and scientific applications. Our work focuses on creating models that bridge the gap between physics and machine learning. These models are reliable and interpretable, providing accurate predictions even when data is limited or noisy.

Explore research areas

Related projects

Landskape

RICO