An Improved Genetic Algorithm and Its Blending Application with Neural Network

Abstract

In order to overcome the limitation such as premature convergence and low global convergence speed of standard genetic algorithm, an improved genetic algorithm named adaptive genetic algorithm simulating human reproduction mode is proposed. The genetic operators of this algorithm include selection operator, help operator, adaptive crossover operator and adaptive mutation operator. The genetic individuals' sex feature, age feature and consanguinity feature are considered. Two individuals with opposite sex can reproduce the next generation if they are distant consanguinity individuals and their age is allowable. Based on this improved genetic algorithm, a thoroughly evolutionary neural network algorithm named IGA-BP algorithm is proposed. In IGA-BP algorithm, genetic algorithm is used firstly to evolve and design the structure, the initial weights and thresholds, the training ratio and momentum factor of neural network roundly. Then, training samples are used to search for the optimal solution by the evolutionary neural network. IGA-BP algorithm was used in pattern recognition example of Gray code. The illustrational results showed that IGA-BP algorithm was better than traditional neural network algorithm in both speed and precision of convergence, and its validity was proved.