TRAINING ARTIFICIAL NEURAL NETWORK BY INVADING ADAPTIVE GENETIC ALGORITHM

A suitable chromosome encoding method, which could correspond with the network one by one, was proposed. The species invasion genetic algorithm was used to train artificial neural networks. In the invading process, the genetic algorithm adjusts adaptively crossing operation and mutation operation. A method based on the average fitness values was proposed to determine the scale of invasion species, and a detailed description of the algorithm steps was given. Finally, the validity and superiority of the algorithm are proved by the experimental results.