Title: Similarity in Visualization und Visual Analytics
Name: Patrick Rüdiger
Phone: 0631 – 205 - 3501
Similarity is a key factor for understanding the process of scientific knowledge gain. Nevertheless, it is commonly used in a fuzzy way and lacks a proper interpretation. Especially visual similarity is mostly reduced to similarity in the underlying representation of data and therefore lacks the possibility of integration into higher levels of information. To understand how visualization works and can be improved on a scientific basis the concept of similarity for visualization needs to be revised.
To come up with a similarity metric than can be applied on high level information, the correlation between similarities in the visual output and the causing effects in the provided data is investigated. The presence of simulation models in varying characteristics gives the opportunity to study this correlation in a way not possible with classical experiments before. The possibility to create artificial models with a predefined set of causing effects and the nearly full control of external conditions allows a precise study in a controlled experimental environment.
Understanding the correlation of similarity between high level information and the causing effects enables a new understanding on how to interpret simulation and experimental results and adds new dimensions for improvements in visualizations based on increasing complex user requirements.
Fig.1 Is the visual similarity correlated with similar changes in the causing effects? To answer this question or to decide if the chosen visualization is helpful a proper understanding of similarity for both cases is needed.