goto contents

Humberto Trimiño Mora

Acceleration of Bayesian model based data analysis

Universität Rostock, 2018

Abstract: Inverse problems for parameter estimation often face a choice between the use of a real-time scheme with strong approximations or rigorous post-processing with explicit uncertainty handling. Plasma physics experiments set a particularly high demand of both and a solution that meets all of these requirements is missing. Standard Bayesian analysis is an excellent tool for the case at hand, with the disadvantage of extensive processing times. This work therefore presents a solution that satisfies the scientific requirements while reducing the need for a speed vs. rigorosity trade-off.

doctoral thesis   free access