特征识别—反馈型神经网络设计

Feature Recognition Based on Elman Feedback Neural Networks

构造一个Elman反馈神经网络来进行模式识别,给出了所构造的Elman反馈神经网络的结构,它相对于BP网络的优势在于它能在有限时间内以任意精度逼近任意函数,在错误概率最小的条件下,使识别的结果尽量与客观事物相符。对二者的识别错误率进行比较,结果表明,反馈神经网络在模式识别的稳定性及真实性上有着BP网络所无法比拟的优势。对如何提高反馈神经网络的辨识精度做了一些探讨。

An Elman neutral network is constructed for model identification.The traditional BP calculate way is lack of satisfication on the stability,study rate and recognition accuracy. In comparison with BP calculate way, it can approach to any function with arbitrary accuracy in limited time and make the identification result agree with the object with minimum mistake rate. The experiment shows that the Elman feedback neural network is better than BP network on the stability and reliability of mode identification.