论文部分内容阅读
提出一种基于梯度下降法的混合进化算法,用于确定径向基函数(RBF)神经网络结构和优化其参数.在进化算法中嵌入梯度下降算子,对每一代中若干个精英个体以一定概率利用梯度下降法进行搜索,以加强算法的局部搜索能力.利用混合进化算法对RBF网络结构和参数同时进行训练和优化,对网络节点数和参数进行混合编码.仿真实验结果表明该RBF网络具有较强的泛化能力.
A hybrid evolutionary algorithm based on gradient descent method was proposed to determine the structure of radial basis function (RBF) neural network and to optimize its parameters. The gradient descent operator was embedded in the evolutionary algorithm, and several elite individuals in each generation Probability to search using gradient descent method to enhance the local search ability of the algorithm.Meanwhile, the hybrid evolutionary algorithm is used to train and optimize the RBF network structure and parameters simultaneously, and the network nodes and parameters are mixed and encoded.The simulation results show that the RBF network has Strong generalization ability.