Optimization of Diagnostic Configurations in the presence of Uncertainty using Bayesian Inference and Optimization

Inferring plasma conditions from experimental data in High Energy Density Physics and Inertial Confinement Fusion experiments is a complex task. The quality of our inferences can be limited by choices we make in configuring the associated diagnostics and uncertainties in a variety of calibration data. Configuring instruments for this purpose is often guided primarily by intuition and fails to account for all known sources of uncertainty that can introduce significant bias and reduce confidence in our inferences. Here we present a method to optimize instrumental configurations using a physics-motivated example with the goal of minimizing bias and uncertainty in inferences while accounting for a variety of unknowns in the experiment and analysis. We show how Bayesian inference and Bayesian Optimization can be combined to provide a powerful and general method for optimizing diagnostic configurations and maximizing our learning from each experiment.