基于分布式神经网络递推预报误差算法的非线性系统建模

采用基于递推预报误差算法的分布式神经网络 结构建立非线性系统模型．子神经网络模型及其连接权值均采用递推预报误差方法来进行训 练，将所有子网络融合得到的分布式神经网络模型在模型精确性和鲁棒性方面有显著地增加 ．该方法较好地应用于复杂非线性动态系统的建模．

DISTRIBUTED NEURAL NETWORKS BASED NONLINEAR SYSTEM MODEL USING PARALLEL RECURSIVE PREDICTION ERROR ALGORITHM

Improved predictions can be obtained by using distributed neural networks (DNNs) instead of an individual better network as usual. New approach to obtain combination weights of multiple networks is derived based on parallel recursive prediction error algorithm. Model accuracy and robustness can be significantly improved by using distributed neural networks. The proposed method has been applied and evaluated for a complex dynamic nonlinear system. Results obtained demonstrate that this approach can improve the performance of neural network based nonlinear models.