While available computational resources seem ever-increasing, the data acquisition in many large-scale scientific problems will remain restricted or expensive also in future due to fundamental physical or economical limitations. This project studies Bayesian optimal experimental design, which aims at maximizing the value of experimental data. We develop methods that guide and accelerate computations needed for large-scale nonlinear inverse problems. The developed techniques are applied to magnetorelaxometry imaging, internal temperature measurements for validating models for iron loss in electric motors, and head imaging by electrical impedance tomography.

2022

2026