Lamarckian GA with genetic supervision

The evolutionary theory advocated by Lamarck (J. Cherfas, 1984), focuses on the inheritance of characteristics acquired for self adaptation to environment. In the domain of the purpose of acquiring adaptive strategies, it is important to make use of the information of experiences through adaptation. Therefore, the Lamarckian mechanism is an effective approach and is expected to augment the power of many kinds of evolving or learning algorithms. We propose the Lamarckian Lookup Table type Genetic Algorithm (LLT-GA). In general, the effectiveness of the characteristics useful for adaptation depends on a class or rather a landscape of problems to be applied. In order to demolish this barrier, the proposed LLT-GA is armed with a control mechanism for acquired characteristics based on a concept of genetic supervision. We discuss first Lamarckian effect and demonstrate that it is dependent on a landscape of a problem. Then, we develop an adaptive evaluation module for Lamarckian evolution with genetic supervision. The simulation results show the effectiveness of the proposed LLT-GA