Improved Genetic Algorithm Based on Critical Self-Organization and Gaussian Memory for Solving Dynamic Optimization Problems

Dynamic components, nonlinear limitation, and multi-objectives are characteristics that we face in the real world. Nowadays, using transmutation algorithms based on biological behaviors is spread e.g. Genetic algorithm. This study tried to design optimization protocol, inspired by Genetic algorithm. Such algorithm tries to keep its complexity and variation in question scope will happen periodically. In other words, this study proposes an optimized Genetic algorithm to solve dynamic optimization problems.
A new self-mutate operator based on the sandhill model was used in this algorithm. a self-mutate operator is a new mutate operator which can predict self-regulated mutation rates based on the sandhill distribution model. This model can match the new cope condition If variations happen periodically. Switching to a new situation is based on memory. One of the issues of using memory is diversity. a density prediction memory with gaussian clusters was used to increase memory diversity. The new method showed better results compare to the other genetic algorithms. Results were compared with the other self-mutate methods. Also, results were tested with on different functions such as royal road, one max, and deceptive. The phase results were much better than the opponent methods. Since the parameters of the proposed method will not increase in comparison with other algorithms, It can be used in real-world applications.