Methods to analyze these data sets have become of great interest in recent years, driven by the availability of cheap, powerful computers and the potential benefits to be gained from insights into these complex data.
The purpose of research in this area is to find methods to support the analyst in this process by developing new techniques to identify and extract meaningful information from these vast amounts of data and to represent the information and relationships identified to the user and so improve the speed, accuracy and completeness of their understanding.
The Information Visualization group is engaged in research into many aspects of this work, ranging from the development of new algorithmic approaches for the extraction of patterns and relationships in data to the representation of these features through interactive computer graphics and sound, as well as other perceptual mechanisms.
Information visualization techniques for analysis of multivariate and time-varying data
This research direction focuses on developing new visual methods for analysis of multivariate, temporal data. One research area is concerned with the development of efficient techniques suitable for detecting complex relationships in very large data sets, utilizing hardware accelerated computations. Research is also conducted on visual quality metrics to quantify loss of information in data abstraction methods. Furthermore, the research concerns the development and user-centered evaluation of methods for reducing visual clutter in the parallel coordinates display.
Interactive visual data mining
This line of research is concerned with one of the main drawbacks of most data mining algorithms. Namely, the fact that most algorithms today operate as a “black box” and the contribution of the end user, that aims to benefit from their results, is often limited to adjusting some initial parameters or settings. The core issue we want to address within this research direction is to increase the interestingness and relevance of data mining results by interactively adding domain knowledge into the process. To accomplish this in our research we are investigating approaches that combine algorithmic data mining with interactive visualization techniques aiming to create a “transparent box” execution model for algorithms, instead of the traditional “black box” one.
Elementary research in the field of visualization and user-centered evaluation. One focus area in the research is development and validation of new evaluation methodology and its applications in visualization. Another focus area is user-centered evaluation of visual methods for high-dimensional data, mainly within information visualization, and evaluation of multi-modal interaction.
Sonification and information visualization
Sonification means that sound is added as a complement to a visual representation to make the representation easier to comprehend or to reveal new relationships in research data. Sometimes visual representations might be hard to comprehend, and the complexity loads the visual cognitive system. By using sound as a complementary modality, it is possible to provide more information and simultaneously ease the interpretation of the visual representation. For example, sonification can reduce visual misinterpretations caused by simultaneous brightness contrast, ease understanding of density levels, or support the perception of different datasets in a visualization.
Musical sonification >>