AGA-BP Algorithm Based On Jumping Gene

BP(Back Propagation)neural network and GA(Genetic Algorithm)are bionic algorithm which simulate the intelligent information processing mechanism of biological objects.It is difficult that only using neural network or genetic algorithm to solve complex classification problems with the in-depth study of two algorithms.In general,BP neural network models have good nonlinear mapping and learning ability when fitting training samples.However,BP neural network may not be a good solution to the classification problem when the network model structure is not fixed,the initial weights and thresholds are uncertain,and the convergence speed is slow.At the same time,the GA has a good search ability for adaptive optimization problems,but it does not have adaptive learning ability similar to neural network,so it is difficult to conduct effective prediction and classification research.In recent years,in order to find the global optimal solution of complex problems,many researchers have tried to combine genetic algorithms with artificial neural network methods.Although the GA-BP learning algorithm has some advantages in solving complex classification problems,it tends to have slow convergence rate and may fall into local extremes.Aiming at the above problems,this paper selects the adaptive GA-BP(AGABP)algorithm,and adds the jumping gene on GA-BP algorithm and AGA-BP algorithm,which is called JG-GA-BP algorithm and the JG-AGA-BP algorithm.These methods are used to deal with the classification problem.The algorithms are proposed to optimize the structural parameters of BP neural network in this paper,and then establish corresponding neural network topology models.In order to verify the classification effect of the learning algorithm after adding the jumping gene,algorithms models are built on random numbers,iris,wine and abalone datasets,and the performances of four algorithms are compared.The experimental results show that the accuracy and convergence speed of AGA-BP algorithms based on jumping gene are improved to some extent.