The goal of this thesis is to demonstrate the development of a learning system based on pattern fitting.
In automation and control engineering, machine learning can provide a big help by discovering opportunities for optimization and providing information about the possible outcomes of a process. This can be beneficial in the handling of complex systems.
The realization of such learning systems can be based on the recognition of certain special patterns or shapes or situations.
The game Gomoku is a usual prototype of learning systems, since it can be precisely described by a small amount of rules, and the decisions and the outcome of the game is deterministic.
In order to be able to quickly and comfortably test the algorithm and also for demonstrational reasons, I created a simulation environment (a Gomoku game) using Matlab.
After summarizing the basics of the theoretical background of the topic and describing the Matlab based simulation environment, I will discuss the actual pattern fitting algorithm and the evaluating and scoring algorithm I have developed. The perfect cooperation of these components makes the system be able to learn and make the right decisions.
I will also describe the Matlab-functions, which realize the algorithms.
Finally I will discuss the achievements.