Keynote Lectures

Abstract
Biomedical Informatics can be defined as “the interdisciplinary field that studies and pursues the effective use of biomedical data, information and knowledge for scientific inquiry, problem solving, and decision making, motivated by efforts to improve human health.” However, professionals in the life sciences are faced with an increasing quantity of highly complex, multi-dimensional and weakly structured data. While researchers in Human-Computer Interaction (HCI) and Information Retrieval/Knowledge Discovery in Databases (IR/KDD) have long been independently working to develop methods that can support expert end users to identify, extract and understand information out of this data, it is obvious that an interdisciplinary approach to bring these two fields closer together can yield synergies in the application of these methods to weakly structured complex medical data sets. The aim is to support end users to learn how to interactively analyse information properties and to visualize the most relevant parts – in order to gain knowledge, and finally wisdom, to support a smarter decision making. The danger is not only to get overwhelmed by increasing masses of data, moreover there is the risk of modelling artifacts.

Abstract
A dataset with M items has 2^M subsets anyone of which may be the one which satisfies our objectives. With a good data display and interactivity our fantastic pattern-recognition can cut great swaths searching through this combinatorial explosion and also extract insights from the visual patterns. These are the core reasons for data visualization. With parallel coordinates the search for relations in multivariate datasets is transformed into a 2-D pattern recognition problem. The foundations are developed interlaced with applications. Guidelines and strategies for knowledge discovery are illustrated on several real datasets (financial, process control, credit-score, intrusion-detection etc) one with hundreds of variables. A geometric classification algorithm is presented and applied to complex datasets. It has low computational complexity providing the classification rule explicitly and visually. The minimal set of variables required to state the rule (features) is found and ordered by their predictive value. Multivariate relations can be modeled as hypersurfaces and used for decision support. A model of a (real) country’s economy reveals sensitivies, impact of constraints, trade-offs and economic sectors unknowingly competing for the same resources. An overview of the methodology provides foundational understanding; learning the patterns corresponding to various multivariaterelations. These patterns are robust in the presence of errors and that is good news for the applications. We stand at the threshold of breaching the gridlock of multidimensional visualization. The parallel coordinates methodology has been applied to collision avoidance and conflict resolution algorithms for air traffic control (3 USA patents), computer vision (1 USA patent), data mining (1 USA patent), optimization, decision support and elsewhere.

Abstract
With the upcoming widespread availability of sensors, more and more applications depend on physical phenomena. Up-to-date real world information is embedded into business processes, in production environments, or in mobile applications, so that such context-aware applications can adapt their behavior to the current situation of their user or environment. Another example are so-called SCADA systems (supervisory control and data acquisition), where complex installations (e.g., energy grids or power plants) are monitored and controlled. In general, data can be managed either by a database management system (DBMS), or directly by the application. The first approach has many advantages: information demands can be declared by queries and kept separately from the application. When the information demand changes, only the query has to be changed (which dramatically decreases software maintenance costs). In addition, a DBMS can optimize the query execution, so that the requested data is retrieved efficiently from the systems. However, if applications depend on real-world data, the amount of data, the update rate, and low latency requirements often prevent the storage in a DBMS. Thus, the data streams from the sensors are managed directly by applications, with all its drawbacks. The goal of data stream management is to provide the same flexibility and data independence for data streams as for stored data. However, despite of performance advantages, DSMS represent no general solution since applications often require the persistent storage of continuous query results as well as a combined processing of current and historical stream data. The performance of integrated database methods, on the other hand, is limited due to the expensive data management of traditional DBMS. A naïve solution would be to federate or integrate both types of systems. However, a closer look shows that such a federation raises many open questions