Research Of Neural Network Structural Optimization Based On Information Entropy

In the application of deep learning,the depth and width of the neural network structure have a great impact on the learning performance of the neural network.The choice of the neural network structure for different training sets is based on experience and multiple experiments to address this inefficiency question,this paper studies an optimization problem for the neural network structure based on the structure of the rule fully connected neural network.First,combining the iterative efficiency of the neural network and its network accuracy,a quantitative method is designed to measure the energy efficiency of the neural network through scores;secondly,the information entropy model and decision tree strategy are used for feature selection and structural adjustment to optimize the neural network candidates.Finally,a heuristic optimization algorithm for neural network structure adjustment based on decision tree is proposed.The experimental results show that the algorithm is applied to the fully connected neural network trained on the iris data set.The energy efficiency score of the network determined by the algorithm is close to the optimal score obtained by the traditional algorithm of training the network one by one,which proves the effectiveness of the method is verified,and the efficiency of the algorithm is verified through comparative experiments.