Skip to main content

Reduced-order Modeling

Traditional modeling techniques can have huge computational costs. This is especially limiting for many- query uncertainty quantification, where uncertainty in model outputs is estimated by repeatedly running the model. Reduced order modeling is a data-driven modeling technique that aims to reduce the computational cost of a model by identifying and exploiting underlying patterns in the data while still maintaining enough accuracy to be meaningful. We are using both reduced order models and traditional high-fidelity models together to reduce the time and cost of many-query uncertainty quantification. This will enable uncertainty quantification to be performed for focused ultrasound modeling, making it a more viable option for treatment planning.