Elmar Steiner

The dissertation explores the optimization of stochastic search through the inspiration drawn from a specific mechanism in the self-organization and balancing of systems of/in honeybee colonies, specifically the decentralized task allocation to individual bees. The study draws parallels between the biological mechanism and meta heuristic algorithmic design, particularly focusing on indicators representing the population's state and search behavior. The research involves measuring whether the algorithm is in an explorative or exploitative phase based on progress in finding promising solutions and characteristics of the search space. This information is then used to dynamically control and balance the algorithm's behavior, particularly in changing perturbation neighborhoods. As information about the change between phases the usage of local optima networks, a fitness landscape abstraction, is proposed. Appropriate control is devised using machine learning techniques. The application is demonstrated in solving various combinatorial optimization problems, known for their difficulty. Additionally, the study integrates applied aspects related to educational systems, particularly in optimizing course selection for students to enhance resource utilization. The developed stochastic procedure proves valuable, especially in scenarios where exact optimization techniques face challenges due to problem size. The research extends beyond engineered organizational problems, showcasing the adaptability and relevance of bio-inspired algorithms in diverse domains.