论文部分内容阅读
为提高径向基函数神经网络的泛化性能,提出一种利用分级偏最小二乘回归方法构造径向基函数神经网络的方法,逐步增加网络中的隐节点数直至达到合适的网络规模,消除了训练数据中存在的多重共线性对网络泛化能力的不利影响.所得径向基函数神经网络的泛化能力比偏最小二乘回归构造的径向基函数神经网络提高了约30%.船舶航向跟踪预测控制仿真验证了该算法的可行性和有效性.
In order to improve the generalization performance of radial basis function neural networks, a method of constructing radial basis function neural networks by using hierarchical partial least squares regression is proposed, and the number of hidden nodes in the network is gradually increased until the appropriate network size is achieved. The multicollinearity existing in the training data has an adverse effect on the generalization ability of the network.The generalized ability of the resulting radial basis function neural network is improved by about 30% compared with the radial basis function neural network constructed by the partial least-squares regression. The heading tracking predictive control simulation proves the feasibility and effectiveness of this algorithm.