The present paper introduces a mutation-based evolutionary algorithm that evolves genes to regulate the developmental timings of phenotypic values. For each generation, an individual in the evolutionary algorithm time-sequentially generates a given number of entire phenotypes before finishing its life. Each gene represents a cycle time of changing probability for determining its corresponding phenotypic value, which is an indicator of developmental timing. In addition, the algorithm has a learning mechanism such that, during the lifetime of an individual, genes representing a long cycle time can change the probability of adaptation more easily than genes representing a short cycle time. Therefore, if the diversity of the genes is maintained, it can be expected that the algorithm provides a different evolution speed to each phenotypic value. The present paper also discusses a new approach to depicting an evolutionary optimization process. An evolutionary optimization process involves the identification of linkage between variables, and therefore, network structures formed using the identified linkage information determine how the evolutionary algorithm solves a given optimization problem. The proposed approach regards an evolutionary optimization process as a change in the network topology that emerges in the process of linkage identification. The simulation results indicate that evolution and learning mediated by the difference in developmental timing helps to sequentially solve hard uniformly-scaled bit optimization problems with linkage between variables.
A Multi-Objective Evolutionary Algorithm Based on KNN-Graph for Traffic Network Attack
