Neural Network Optimization Based On Beetle Antennae Search Algorithm And Its Application In Pattern Classification

With the development of computers and the enhancement of computing power,artificial neural networks,as an important representative member of pattern recognition,have become a research hotspot.Error Back Propagation Training(BP)neural network based on gradient descent(GD)algorithm is one of the most mature artificial neural network models in application direction.However,the traditional BP neural network model based on GD algorithm has shortcomings.For example,the convergence speed of the algorithm is slow,it is easy to fall into the local optimum,and the optimal neural network structure cannot be determined.The beetle antennae search(BAS)algorithm has a simple structure and few parameters.In this paper,We use the BAS algorithm in the BP neural network model and improve the model.This makes up for the shortcomings of the BP neural network model based on the GD algorithm and the slow convergence of the algorithm and the inability to determine the optimal neural network structure.The main research content of this paper can be summarized as follows:First,this paper proposes a new model combining the BAS algorithm with BP neural network for the slow convergence of BP neural network model based on GD algorithm,termed BASNNC.The BAS algorithm is used to optimize the weight of the BP neural network.At this point,the BP neural network does not need to use GD algorithm to optimize the weight.The BAS algorithm is a heuristic search algorithm.Compared with other population search algorithms such as genetic algorithms,the biggest advantage of BAS algorithm is that only a single individual is needed,and the calculation amount is greatly reduced.We performed two test functions and a classification application on nine UCI datasets.The experimental results show that the proposed algorithm achieves the same classification accuracy as the BP neural network classification model based on the GD algorithm.The algorithm not only reduces the computational complexity,but also speeds up the convergence of the algorithm.Furthermore,for the BP neural network model based on GD algorithm,the challenge of determining the optimal neural network structure cannot be determined.We propose an improved algorithm of BAS-NAS for the optimization of neural network structure.An improved BASNNC algorithm(i BASNNC)is proposed to expand the search for parameters in neural networks.The network structure search in this paper is mainly to optimize the number of neuron nodes in the hidden layer in the three-layer neural network(i.e.,the input layer,the hidden layer and the output layer).At present,the most commonly used method for determining the number of hidden layer neurons is the trial and error method,which is obtained according to experience and multiple parameter debugging.In this paper,the BAS-NAS algorithm combined with the i BASNNC algorithm is used to adaptively obtain better parameter combinations,thus forming a better neural network structure.The MATLAB simulation experiment proves that the classification accuracy of the improved neural network model has been increased.