Beschreibung:
The simulation of gravitational microlensing light curves is a time-consuming effort, but essential for confirming potential planetary microlensing events. In this thesis a point-wise binary lens model was developed and parallelized on Graphics Processing Units (GPUs). As most graphic cards are processing single-precision floating point numbers, special care was taken to understand and control the numerical accuracy. In order to increase the efficiency of the fitting process, information theoretical concepts have been exploited. Based on the insight, that observations differently contribute to the total information content, subsets of the light curves are selected according to the current state of a Markov Chain. Furthermore, the information content of a point lens light curve was used in order to optimize the sampling of a light curve. In a further major project, follow-up observations were carried out as part of the MiNDSTEp consortium and a data reduction pipeline was prepared ...