Comparing genetic operators with Gaussian mutations in simulated evolutionary processes using linear systems

Abstract. Evolutionary optimization has been proposed as a method to generate machine learning through automated discovery. Specific genetic operations (e.g. crossover and inversion) have been proposed to mutate the structure that encodes expressed behavior. The efficiency of these operations is evaluated in a series of experiments aimed at solving linear systems of equations. The results indicate that these genetic operators do not compare favorably with more simple random mutation. _____________________________________________________________________________