Construction And Application Of Random Wavelet Polynomial Neural Network

Random neural networks combine the advantages of random theory and neural networks,and their powerful generalization and self-learning abilities have achieved extraordinary results in security prevention and control,pattern recognition,medical images and other fields.However,traditional random neural networks also have some shortcomings,especially the disadvantage that the generalization ability does not meet the practical needs when building multi-dimensional complex models and the tendency to fall into the overfitting dilemma.In this paper,we propose a random wavelet polynomial neural network to overcome the above problems,and the main innovations are as follows:1.A random polynomial neural network is proposed.The model has the following outstanding advantages: generalization ability,and which can well reflect the higherorder nonlinear relationship between the input and output data.The random polynomial neural network combined with the idea of ensemble learning,random polynomial neural network uses random resampling technology to generate multiple sub data sets,and uses the generated sub data sets to construct multiple polynomial neural networks,and then aggregates a single polynomial neural network to form a random polynomial neural network.The least square method is used to estimate the coefficients of polynomial function,which reduces the amount of calculation of the model and improves the generalization ability of random polynomial neural network.2.A random wavelet polynomial neural network optimized based on particle swarm algorithm is proposed.The model is improved on the basis of random polynomial neural network,which has better classification performance.In addition,the model can better process time domain signal data.The parameters of the model are estimated using L2 regularized least squares method.In addition,the particle swarm optimization algorithm is used to optimize the structural and input parameters of the model,which further improves the generalization capability of the proposed model.3.The proposed neural network is applied to face recognition and atrial fibrillation diagnosis.A face recognition model based on stochastic polynomial neural network is proposed.Firstly,the local binary pattern algorithm is used to extract the features of human face,and then the principal component analysis method is used to reduce the dimensionality of face features,and finally the random polynomial neural network is used to classify faces.In addition,a random wavelet polynomial neural network optimized based on particle swarm algorithm is proposed for atrial fibrillation diagnosis.The atrial fibrillation signal is denoised,and then the denoised data is feature extracted,and finally input to the proposed model for diagnosis.The experiments show that the classification correct rate of the proposed model is better than that of some existing models.