Research On BP Neural Network Optimization Based On Inverse-time Chaotic Coyote Optimization Algorithm

The performance of biomimetic optimization algorithm and the efficiency of neural network training methods have always been the focus of machine learning research.As optimization problems in science become more and more complex,bionic optimization algorithms are constantly proposed and improved,but there is still room for improvement in terms of convergence and other aspects.Traditional neural network training methods based on stochastic gradient descent have been unable to solve these complex optimization problems.How to optimize neural network parameters and explore high-performance neural network training methods has become one of the important problems in machine learning.Aiming at the above problems,this paper proposes a BP neural network model based on the improved coyote optimization algorithm and the parameter optimization of BP neural network.Firstly,to solve the problems of weak performance and low diversity of coyote optimization algorithm,a chaotic coyote optimization algorithm(ICCOA)based on inverse time attenuation operator was proposed in order to better play the performance advantage of coyote optimization algorithm in solving complex optimization problems.On the one hand,in the process of individual iteration,the inverse time attenuation weight factor is added to keep the global search and local development ability in balance and improve the search speed of the algorithm.On the other hand,the chaotic disturbance mechanism based on Tent chaotic mapping is added to generate new individuals by mapping some poor individuals in the population,thus increasing the population diversity.In order to verify the optimization ability of ICCOA algorithm,function optimization tests were carried out in 10,30 and 100 dimensions respectively,and compared with five optimization algorithms.The experimental results show that ICCOA algorithm has good optimization performance.Secondly,to solve the problem of low efficiency of BP neural network,the unique evolutionary strategy of improved coyote population was introduced into the parameter optimization process of BP neural network.In this paper,a parameter optimization method of BP neural network based on inverse time-limit chaotic coyote optimization algorithm was proposed as a new neural network model(ICCOABP).The ICCOABP algorithm is applied to the pattern recognition task and numerical prediction task of machine learning together with the standard neural network and BP neural network parameter optimization method based on genetic algorithm.The experimental results show that the ICCOABP algorithm has high efficiency.The new BP neural network model adopts the improved coyote optimization algorithm as the optimizer,which greatly improves the prediction performance of the model and has great theoretical and practical significance.The paper has 24 pictures,6 tables,and 73 references.