These days, data analysis is no longer this long-winded process that churns on the data until hours later a final result has been computed. With current hardware speedups and new computational paradigms like Progressive Visual Analytics and Approximate Query Processing, a computational analysis run by the machine and an interactive visual exploration performed by the user can be carried out side by side with results of the computation and of the exploration informing each other. This way, a well-run visual data analysis can turn almost into a synchronous collaboration scenario between human analyst and machine.
This PhD project sets out to investigate how to best facilitate this collaboration by developing novel visual interaction techniques for such a closely intertwined computational data analysis and interactive visual data exploration. As computations are used quite differently alongside a running analysis – from silent background processes that once deployed only chime in with new results every now and then (e.g., progressive probes), to visually very present foreground processes that require active direction and steering by the user (e.g., computational smart lenses) different techniques to interact with them will need to be explored.
Research questions to pursue in this project are:
- Which metaphors are appropriate for the otherwise invisible computational process and to which kinds of visualization and interaction do they lead? (e.g., computational analysis as a tool to be used by the analyst, as a resource to be distributed between different analytical objectives, data subspaces, or levels of granularity, as well as an independent helper that can be tasked by the user and reports back when needed)
- Which progressive processes – data transmissions (e.g., streaming data or progressive data loading), analytic computations (e.g., progressive clustering or density estimation), and visualizations (e.g., progressive rendering or iterative layouting) – can be shown and interacted with in which ways?
- Can and should the computational steering be coupled with the interactive visual exploration? For example, a zoom-in on a region of interest in the view could at the same time focus the running computation on that particular data subspace as well. Or should both complement each other, so that if the user looks at some specific data subset, the computational process monitors the remainder of the data for out-of-sight changes?
Applicants seeking further information are invited to visit: http://phd.scitech.au.dk/for-applicants ... analytics/
Prof. Dr.-Ing. Hans-Jörg Schulz
Department of Computer Science
University of Aarhus
8200 Aarhus N, Denmark