Automatische Konfiguration von Prüf- und Diagnosesystemen

The method involves evaluating a productivity of a configuration by using random samples. The configuration of the pattern recognition system is modified in direction of a more suitable configuration. It is modified with the help of optimization strategies, such as linear or non-linear optimization, genetic algorithms or genetic programming. The steps are repeated to achieve an iteration to a maximum productivity. Alternatively a learning phase is included in the evaluation, in which a classifier is adjusted through automatically selected learning patterns.