Institut für Programmstrukturen und Datenorganisation


ABBRECHEN || neuen Vortrag anmelden | Vorträge durchsuchen
Monat zurück | Vorträge dieser Monat | Monat vor

Fr, 03.03.2017 Herr Zengchao Geng (Masterthesis) ID: 100276
Co-Evolution of Metamodels and Database-stored Instances
Betreuer: Jörg Henß
Model-Driven Engineering (MDE) changes the focus of the software development from coding to modeling. As the large models cannot be loaded into memory, they should be persisted into database. Changing requirements and detection of bugs often lead to new metamodel versions. When the metamodel evolves, database-stored instances that previously conformed to the metamodel might no longer conform to the new structures and constraints designed by the changed metamodel. Manually migrating the existing instances in database is tedious and error-prone. The concept of operations mapping and translation strategies are proposed in this thesis. Operations mapping means the mapping between the operations of metamodel evolution and the operations of database refactoring. Through the investigation of the operations mapping, the translation strategies are generated to translate the history file of Edapt into ChangeLog file of Liquibase and to make sure the automatically co-evolution of metamodel and database-stored instances.

Fr, 10.03.2017 Herr Andreas Schatz (Proposal) ID: 100277
Studying the Placement of Outliers Hidden in Subspaces to Prevent Catastrophic Failures or Sabotage
Betreuer: Georg Steinbuß
In this proposal talk we will discuss the problem of generating realistic hidden outliers. High dimmensional data poses problems in detecting outliers, because of the curse of dimensionality. One way to circumvent this is to restrict attention to lower dimensional projections. In some of these subpsaces the outlier can be detected, while in others it will appear like an ordinary data point and is thus hidden. We will also discuss how expert constraints on these outliers can be taken into account.

Fr, 31.03.2017 Herr Francesco Di Lillo (Masterthesis) ID: 100288
Generation of Synthetic Energy-Consumption Data in Industrial Settings
Betreuer: Holger Trittenbach
The use of synthetic data is an increasing trend in all research fields thanks to the possibility to completely control the generating process, the statistical characteristics and quantity of these data. A field that makes extensive use of synthetic data is industry, since in many cases real word energy data are not fully known, not accessible for privacy reasons or cost and they do not always clearly show behaviour of interest required for research purpose, while using a synthetic data generator it is possible to create datasets that are a representative versions of the original data and, at the same time, follow all the required constraints. The presentation describes the process steps behind the development, evaluation and validation of all the different models used in this research as basis for building a synthetic data generator in order to address the challenge to generate realistic energy consumption data for industrial settings.