Fourth SFB 901 Sem­in­ar in SS 14

On July 2, 2014, the 4th SFB 901 seminar in the summer semester will take place.
                                                                                                                           

We will have two talks in this session:

16:00-16:45 (A3) 
Title: Selfish resource sharing and analyzing complex, repeated strategic behavior
Speaker: Alexander Skopalik


Abstract:

In the first part of this talk we introduce the concept of budget games. Players choose a set of tasks and each task has a certain demand on resources. Each resource has a budget. If the budget is not enough to satisfy the sum of all demands, it has to be shared between the tasks. We study the complexity of the optimal solution as well as existence, complexity and quality of equilibria as well as the effect of simultaneous strategy changes by multiple players. In the second part we present a novel approach to analyze strategic situations in which many players repeatedly face strategic decisions. Such situations arise, for example, in OTF-markets on which OTF-providers have to choose between subsets of service providers and service providers may have the choice of effort (time, money, computational power,...) they use do deliver a service. To design such a market and provide tools like reputation or recommendation systems it is vital to understand such markets. We present a simulation-based approach that nicely complements theoretical and experimental techniques. We discuss future directions, challenges and limitations of our technique.

17:00-17:45 (C3) 
Title: Ontology-based Representation of Optimization Models
Speaker: Florian Stapel

Abstract:

This talk is concerned with a new approach for the representation and processing of abstract optimization models in the generation and execution of distributed Decision Support Systems (DSS).
As typical for an application domain to the CRC, current DSS developments focus on service-oriented architectures. Typical problems to deal with are the integration of optimization models and data, as well as the composition task itself.
Our approach represents model and data structure as instance knowledge of ontologies for optimization and application domains, thereby capturing necessary semantics. The ontologies can, e.g., be used for the (semi-)automated generation of adaptors in between services for model instantiation, solution and postprocessing, as well as for the support of typical search and service composition tasks.
Besides that, the composition of abstract optimization models out of single goal and constraint entities itself is a key-feature we address. To that purpose, we present an  architecture for decoupling the model expression structure from such model entities and data model conceptualizations. By this separation, ontology constituents can be merged into whole models, thereby providing a high degree of flexibility and reusability. By exploiting ontology and XML technology, conflicts can be avoided and model validation can take place. In an OTF-scenario we discuss, how (partial) models can be wrapped into services with accompanying service descriptions in SSL, such that the model composition task can also be lifted to the OTF service composition level. We demonstrate our approach on a network flow example.