Visual support for the understanding of simulation processes

Abstract

Current visualization systems are typically based on the concept of interactive post-processing. This decoupling of data visualization from the process of data generation offers a flexible application of visualization tools. It can also lead, however, to information loss in the visualization. Therefore, a combination of the visualization of the data generating process with the visualization of the produced data offers significant support for the understanding of the abstract data sets as well as the underlying process. Due to the application-specific characteristics of data generating processes, the task requires tailored visualization concepts.In this work, we focus on the application field of simulating biochemical reaction networks as discrete-event systems. These stochastic processes generate multi-run and multivariate timeseries, which are analyzed and compared on three different process levels: model, experiment, and the level of multi-run simulation data, each associated with a broad range of analysis goals. To meet these challenging characteristics, we present visualization concepts specifically tailored to all three process levels. The fundament of all three visualization concepts is a compact view that relates the multirun simulation data to the characteristics of the model structure and the experiment. The view provides the visualization at the experiment level. The visualization at the model level coordinates multiple instances of this view for the comparison of experiments. At the level of multi-run simulation data, the views gives an overview on the data, which can be analyzed in detail in time-series views suited for the analysis goals. Although we derive our visualization concepts for one concrete simulation process, our general concept of tailoring the visualization concepts to process levels is generally applicable for the visualization of simulation processes.

Visual Parameter Space Analysis: A Conceptual Framework

“…Such models can come in very different forms. For instance, stochastic models [14,74] yield different outputs for repeated runs with the same input parameter setting. On the other hand, deterministic models [18,69] produce the same outputs whenever run with the same parameter setting.…”

Integrating Open Data on Cancer in Support to Tumor Growth Analysis

“…Domain-as well as application-specifics need to be taken into account to choose the right visualization tool for supporting search and exploration in general data exploration [4][5][6]. In previous work, approaches for discovery of relevant data in research data repositories based on exploration and visual querying have been proposed.…”

A Visual Analysis Concept for the Validation of Geoscientific Simulation Models

“…Visual support in the modeling and simulation cycle in other domains has been presented by Matkovic et al [24,23] for engineering applications and Unger and Schumann [36] and Schulz et al [29] for cell biology. While being valuable for the respective analysis objectives, the presented approaches do not support data validation in Earth system modeling as intended by our application -which requires a simultaneous consideration of various parameterizations and outputs of a model ensemble, observation data, and computationally determined goodness of fit values.…”