Abstract
Computational simulation became popular not only in the fields of physics or chemistry, but also medical, engineering, finance, and so on. Discovery of the best simulation parameters is an issue in order to fit the result of simulations to real experimental results. This paper presents a system that realizes interactive optimization of such simulation parameters with a visualization technique and distributed computing architecture. The system iteratively executes simulations with gradually varying parameters, and then records the errors of simulation results. Then the system visualizes the distribution of the errors by Heiankyo-View, a novel hierarchical data visualization technique. We assume that there are multiple local-minimum errors in the distribution, but the visualization results help users to interactively choice the best parameters, or continue simulations with interactively varying the parameters.
