Research On Improved Cascade Correlation Neural Network Model Based On Group Intelligence Optimization Algorithm

Cascade Correlation learning algorithm is a special supervised learning algorithm for artificial neural network architecture.Unlike the traditional neural network algorithm which only adjusts the weights in the fixed topology network,it can automatically adjust the network structure and weights in the training process according to the training task and process state to complete the training goal.The optimization target of the new hidden unit,the selection of the activation function and the optimization result are key elements in the Cascade Correlation learning algorithm.The optimization algorithm in the traditional neural network has the disadvantages of single optimization goal,slow convergence speed,and easy to fall into local,which cannot fully meet the key elements in the Cascade Correlation learning algorithm.In order to solve the above problems,we design appropriate group intelligence optimization algorithms which are applied in the weight training of Cascade Correlation neural network.Experiments show that the proposed algorithm improves the fitting ability of Cascade Correlation neural network.It reduces the required number of hidden units as well as the network depth,and optimizes the network structure.The academic contributions of this paper are as follows:(1)A single objective group intelligent optimization algorithm j DE-B is proposed suitably for the weight training of Cascaded Correlation neural network.It improves the ability of Cascade Correlation neural network fitting ability.(2)The limitation of a hidden unit as a separate hidden layer fitting ability is proposed during the training process of the Cascade Correlation neural network.New network structure adjustment scheme and multi-objective training method are proposed.(3)A multi-objective group intelligent optimization algorithm MOEA-T is proposed suitably for the weight training of Cascade Correlation neural network,which reduces the depth of Cascade Correlation neural network.