基于Bayes分析的组合多神经网络函数逼近模型

A Combined Multiple Neural Network Model for Function Approximation Based on a Bayesian Analysis

讨论了复杂函数的多神经网络逼近模型的建立方法．针对映射网络泛化能力差以及结构难以确定等问题，提出了一种基于Ｂａｙｅｓ分析的组合多神经网络建模方法．仿真研究表明，这种建模方法提高了神经网络模型的逼近能力。

The multiple neural metowrk paradigms for complicated function modelling are discussed. A combined multiple neural network paradigm and the relevant modelling algorithm based on a Bayesian analysis to overcome the difficulty met in the development of neural network approximation models are proposed. Experimental results show that the Bayesian analysis based multiple neural network model can improve the approximation capability in the complicated function modelling.