Parametric optimization with evolutionary strategies in particle physics

Evolutionary strategies provide a generic means to find optimum solutions for large-scale numerical problems. The EVA library, an implementation of parallel Evolutionary Algorithms, adds the possibility to run the optimization in parallel on devices ranging from multi-processor machines over clusters to the world wide grid. This paper discusses the application of the EVA library to two distinct computationally expensive physics problems—the optimization of particle signals in histograms and the minimization procedure involved in analysing Dalitz plots.