Automated parameter optimization for LPT measurements using multi-objective evolutionary algorithms

Abstract

The selection of suitable evaluation parameters for Iterative Particle Reconstruction (IPR) is an important aspect when processing particle tracking experiments. Such parameters generally need to be selected individually for each experiment setup and require some experience to adjust. We propose an optimization procedure based on evolutionary algorithms tailored specifically to the IPR approach that can select suitable parameters automatically. Besides assisting with routine evaluations such an optimization can be especially useful when investigating novel components or improvements to the standard IPR approach. Existing experience might no longer apply in such situations due to the complex interplay of individual components in the iterative algorithm. In such situations an optimized set of parameters can serve as a more objective evaluation of any modifications as opposed to a manual tuning of parameters which might miss unexpected synergies and effects. We implement such an optimization approach and utilize it to investigate the effect of an improved peak detection scheme within an IPR evaluation at high seeding densities. In addition to the synthetic experiments we also present a pathway towards application on real-world experiment data.