The still quite young research field Visual Analytics creates visual solutions for data-based questions. We are concentrating on one of the most important future topics in this area: How we can best visualize very large data in order to make it analyzable and further usable. This is of enormous relevance for many fields of research and application, which is why our research work is also very diverse: We are working on effective solutions for the analysis of high-dimensional time-based and extremely large data as well as on the removal of the last barriers between user and data using immersive analytics. We are researching optimization and expansion possibilities of web- and cloud-based solutions for visual analysis as well as in the fields of visual analytics for ensemble simulations and visual parameter space analysis. In addition, we develop methods to equip visual analytics with meaningful statistical control mechanisms to make visualizations more "trustworthy".
High-dimensional time-based data, such as those generated in the Internet-of-Things (IoT) area, pose many challenges for research due to their sheer quantity and size. We develop new visualization techniques or make existing ones more effective to make this data visible and readable.
In the area of Immersive Analytics, we are concerned with how we can take analytical processes further and detach them from the traditional desktop computer. In order to build more efficient bridges between the user and his data, we mainly ask ourselves technical questions: In which ways can data be presented? How can interaction with data beyond keyboard and computer mouse still work and for what purposes does it make sense?
As more and more data accumulates in industry, which is also more often stored centrally, the importance of web- and cloud-based solutions continues to grow. Such solutions are also pleasant for users: everything is in one place and is accessible at best through intuitive interfaces. We create real solutions for the industry and optimize network traffic or research on particularly user-friendly interaction systems.
The fact that not all data is the same is another topic that keeps us very busy in the field of visual analytics. Because not all results made visible by visual analysis are statistically meaningful data. To address this problem, we are researching innovative visual methods that check the relevance of data implemented in existing visual analytics tools. By equipping Visual Analytics with statistical control mechanisms, we can also contribute to making "trustworthy" visualizations.
Since 2005, we have been working intensively on future-oriented applications of Visual Analytics for Ensemble simulations. In cooperation with our industry partner AVL, we have been able to develop innovative visualization methods and completely new explorative analysis methods for simulation results in the automotive industry, which can also be translated to other industries and research fields that work with a large number of data and parameters.
Visual parameter space analysis is of great importance when working with simulation models, because there is an enormous amount of available input and output data in all its complexity that needs to be made understandable. We develop tools to accelerate simulation processes by means of these analyses: e.g. by using visual analysis of the respective parameters to provide information in advance about the usefulness or success of the simulation result.
Modern imaging techniques help to better understand biological processes on the basis of huge amounts of data. In the life sciences, visual-analytical methods for the investigation of large amounts of datacan therefore help to improve the understanding of the data and optimise analysis workflows. We not only focus on the design of applications that enable the most effective answers to scientific questions, but also on the associated data preparation, statistical analyses and indexing as well as the development of high-performance data structures.
Designing decision trees requires both domain knowledge and knowledge of the purpose of the model, which is why this process is very difficult to automate. TreePOD successfully takes a new approach and automatically generates several decision tree candidates with different parameter settings, allowing the most appropriate decision tree to be selected. TreePOD can be used for a variety of tasks and can also be integrated into existing workflows for creating and selecting decision trees. Click here to watch a video about TreePOD.
WeightLifter is a novel, interactive visualization technique that facilitates the exploration of weighting spaces with up to ten criteria. Our technique makes it possible to better understand the sensitivity of a decision to changes in weights, to efficiently locate weight ranges where a particular solution ranks high, and to filter out solutions that are not ranked high enough for a plausible combination of weights. Click here to watch a video about Weightlifter.
Intelligent simulations to improve interactive lighting concepts
In our research work within the HILITE and Sharc projects we develop tools and methods that optimize simulation processes for interactive lighting design. Since the processing of enormous amounts of data from different sources - quality of light sources, angles, spatial situation, etc. - must be taken into account, we use principles of visual parameter space analysis to accelerate the workflow massively and to calculate automatically generated suggestions for improvement of lighting situations. Such research projects emphasize the necessity of the interaction between spatial and abstract data worlds to make such and similar analyses possible in the first place.
We successfully apply Visual Analytics methods to help physicians and scientists to explore complex, heterogeneous data in the fields of Medicine and Life Sciences. These data come, for example, from studies on patient cohorts and observations, behavioral data from animal experiments, genetic data, but also, for example, from large collections of spatial measurement, image and network data on the brain. Our Visual Analytics solutions help when initial hypotheses need to be developed based on large amounts of data, but also when there is not enough data available to develop meaningful statistical results and models. Examples of our work include a Visual Analytics framework that supports the search for biomarkers for very rare cancers in children (see also: Visual Analytics and Data Science for the Healthcare System and Medical Research and our research project VISIOMICS) and comprehensive data management, data mining and Visual Analytics solutions for the Neurosciences, some of which are used worldwide (see also: Neuroscience - Visual Computing, Data Science and Big Data and Brain*).
Training AI algorithms requires a great amount of data. However, raw data often contains sensitive information. Homomorphic encryption offers a solution for secure machine learning - with protected sensitive data.
Renewable energies, such as hydropower, are an essential backbone for a climate-friendly future. To enable better real-time monitoring and thus maintenance of hydropower infrastructure, innovative digital hydropower twins are being developed in this project.
The Rail4Future project is focusing on the design of a digital rail system for the future. To this end, a novel and fully virtual validation platform for large-scale simulations of entire rail lines is being developed to increase the efficiency of existing rail infrastructure.
The aim of the application project IVC Multi is to research novel intelligent visual computing methods supporting decision-making in automotive industry, medicine, and life sciences based on ensembles of heterogeneous, multi-scale and/or multi-temporal data.
The strategic project of the area is the organizational and scientific hub for the realization of the area-wide intelligent visual computing approach for analytics and modelling based on ensembles of dense grid-based data, derived data, and digital embedding.
The primary goal of project INGRESS is to accelerate and improve the process of data scientists working with Industry 4.0 and Internet of Things (IoT) data, by enabling a closer integration of visual analysis into the existing workflows.
The RAILING project deals with the research and development of interactive, scalable and trust-building visualization and analysis tools for the exploration of time-dependent and complex data.
Within the project Larvalbrain 2.0, a dynamic multi-scale multi-level atlas and data collection of structural, molecular, physiological, and behavioral results of Drosophila melanogaster larvae will be established.
To take full advantage of data from automotive simulations and measurements, we combine interactive and automatic Visual Computing methods to find intuitive, efficient and effective solutions that are applicable in the daily routine of engineers.
The goal of the applied research project En2VA (“Visual Analytics for Energy and Engineering Applications”) is to increase the efficiency and the quality of advanced analytics for high-dimensional data from manufacturing, engineering, and the energy sector.
Novel visual analysis technologies for high-dimensional data in automotive engineering, industrial manufacturing and the energy sector
Seamless visual analysis of data involving 3D geometry, relational information and multivariate attributes.
DEXHELPP develops new methods to support analysis, planning and control in health care by combining decision analysis, data security, data management, statistics, mathematical modelling, simulation and visual analysis.
Visual Analytics for Modeling and Simulation: Improvement of simulation setup and design scenarios with tools and methods of Visual Analytics.