Research And Application Of Unconstrained Optimization Method Based On BP Neural Network

The BP neural network model is one of the most widely used multi-layer feedforward neural networks.It has the advantages of adaptive,autonomous learning,strong nonlinear mapping ability and strong generalization ability.The theory proves that when the number of neurons in the hidden layer is sufficient,the BP neural network can fit various complex nonlinear functions with only three layers of structure.In practical applications,it is desirable to obtain an input value so that the output obtains an extreme value,that is,an optimization problem based on BP neural network.The existing literature generally uses the gradient method or quasi-Newton method to solve the unconstrained optimization problem based on BP neural network.However,the gradient method and quasi-Newton method are prone to local convergence in the process of solving,which falls into local optimum and cannot obtain the global optimal solution..The genetic algorithm is an adaptive global optimization probability search algorithm,which can perform multi-point search in the solution space at the same time in the search process,which has high calculation speed and can avoid the accuracy of the algorithm falling into the local optimal guarantee calculation.Therefore,the genetic algorithm based on BP neural network unconstrained optimization problem is proposed for the first time,and compared with the traditional gradient method and quasi-Newton method.The results show that the genetic algorithm is used to solve the unconstrained optimization problem based on BP neural network.The effect is better and the calculation speed is faster.On this basis,the genetic algorithm based on BP neural network unconstrained optimization problem proposed in this paper optimizes soybean planting density and fertilization amount in Heilongjiang Province.The main research work is as follows:(1)In this paper,the principle and structure of the standard BP neural network are deeply studied and the BP algorithm is deduced.The gradient method,quasi-Newton method and algorithm steps of the unconstrained optimization problem based on BP neural network are analyzed.It was found that it fell into local optimum during the solution process.(2)In order to overcome the problem that gradient method and quasi-Newton method are easy to fall into local optimum when solving BP-based neural network unconstrained optimization problem,a genetic algorithm based on BP neural network unconstrained optimization problem is proposed.The unconstrained optimization method of BP neural network is tested and compared.The results show that the three unconstrained optimization methods based on BP neural network can obtain the approximate optimal solution when inputting the BP neural network.However,the genetic algorithm based on the BP neural network unconstrained optimization problem has faster speed,higher computational efficiency,and avoids falling into local extreme points for the objective function with multiple local extremum points.(3)The genetic algorithm based on BP neural network unconstrained optimization problem proposed in this paper optimizes soybean planting density and fertilization amount in Heilongjiang Province.