Neural network structure optimization based on improved genetic algorithm

For structural optimization of neural networks, i.e., the challenging problem to determine the number of hidden layers and the number of neurons, we propose a structural optimization algorithm based on an improved genetic algorithm (IGA). The proposed algorithm is then employed to approximate nonlinear function y=e^-(x-1)2+e^-(x+1)2 in MATLAB. Extensive simulation demonstrates that the proposed optimization algorithm is efficient, improves adaptability and generalization ability of neural networks, and holds rapid global convergence.