Research on Optimization of GA-BP Algorithm Based on LM

Neural network is an algorithm inspired by the human brain to perform specific tasks. Integrating the GA-BP algorithm is beneficial to improve the convergence speed and accuracy of the algorithm, but there are still shortcomings such as slow convergence speed and excessive error accuracy. Combined with the characteristics of the LM algorithm, the shortcomings of the GA-BP algorithm are improved. The fitting of the third power function model and the comparison of experimental results verify that the performance of this integrated algorithm model is significantly optimized compared to the original GA-BP algorithm. First, use the GA algorithm to find the optimal individual to obtain the initial weight and threshold of the BP neural network by decoding; then use the LM algorithm to optimize the GA-BP algorithm model, so as to effectively avoid the network training falling into a local minimum or failing to converge. This integrated algorithm obviously improves the network convergence speed, improves the accuracy of model training and the generalization ability of the model network.