Studentenvorträge zu Projekt- und Abschlussarbeiten  

 

Fabian Knopf : GMF-based layouting for the MIO Workbench

Forschungspraktikum - betreut von Dr. Philip Mayer

The MIO Workbench, a formal verification tool for modal input/output
automata, lacks advanced layouting and editing features. This thesis
examines these shortcomings and provides a concept for the conversion
to a GMF-based layouting solution. It details the necessary
frameworks, course of action and problems during the conversion and
concludes with an in-depth examination of selected components of the
final GMF-based Workbench.

-----------------------------------------------------------------------

 

Korbinian Moßandl: JMobWrite - Differentielle Synchronisation in GWT

Projektarbeit - betreut von Dr. Axel Rauschmayer

This document deals with the topic of collaborative text editing, or in other words different people working on the same text at the same time. Therefore the technique of differential synchronisation will be evaluated, which is used in the collaborative text editor MobWrite. Another part of this publication is the implementation of JMobwrite, a port of MobWrite in GWT. At first there will be a basic explanation of the main features of MobWrite, as well as its partial conceptions. After that GWT will be covered in detail along with its advantages and shortcomings. Furthermore the technique of „server push“ will be illustrated and afterwards the focus will be shifted to JMobwrite, the porting of MobWrite. Here the focus will be JMobWrite's concepts and implementation. At the end this ddocument will talk about related work on this topic and an outlook for further improvements will be given.

-----------------------------------------------------------------------

 

Orlin Gueorguiev: Implementation of A Recommendation System in A Social Network

Forschungspraktikum - betreut von Gefei Zhang

One of the main features of social networks is the possibility to analyze the different users, and what they like and dislike. Many features are built on those observations many other features can be build, such as a powerful recommendation system.

This work extended a social web application by the support for automatic neural network based recommendation based on tags. The web application allows the user to specify different publications, which have different set of tags. Each publication has a rating. Based on these information, our recommendation systems suggests new publications to the user, which may be of an interest. The user then has the possibility to rate the recommendations. This information is used to improve the accuracy of the recommendation system. The system is implemented using the popular Ruby on Rails web application framework.

The neural network functionality is delivered from the Ruby Fann gem. The rating system is achieved through the usage of ajaxful-rating gem. The glue between those three modules is implemented in Ruby.

-----------------------------------------------------------------------

 

Guilin Yang: A Distributed Soft Constraint Solver

Projektarbeit - betreut von Gefei Zhang

Constraint programming is a declarative programming paradigm, in which a set of  constraints are used to restrict an object, and a constraint solver efficiently traverses the whole solution space. In the classical constraint programming, the programmer can only formulate hard conditions, such conditions that must be met by the solution. Soft constraint programming generalizes this approach and enables the definition of soft constraints and a soft constraint is a measure of the quality of a solution: the better a solution, the higher the soft constraint value. This way, soft constraints define relations for comparison of possible solutions. An individual, soft constraint therefore defines a “dimension”, in which the solution can be optimized. For example, when buying a computer the performance of the CPU may be taken as a Soft-Constraint. In many cases it is necessary to calculate a solution of the soft constraint system and thus to find a compromise of all dimensions. For instance we need a computer with maximized performance and minimized price and minimized consumption. Since the search tree may be very large, solving soft constraint problems distributedly would be highly desirable. In this thesis, we develop a distributed Soft-Constraint Solver with Branch and Bound Algorithm based on the well-known framework JPPF.