含体液因子影响的神经网络仿真

Simulation of Neural Network Affected by Humor Factor

调节人的机体活动的方式除了神经调节还有体液调节等方式。本文将体液调节与人工神经网络相结合,对一种结合体液调节的人工神经网络机制进行了仿真(相当于内环境可变的神经网络系统)。在不考虑神经系统对体液系统反作用时,对BP神经网络的受体液环境参数影响的阈值参数学习公式进行了推导。接着,考虑到神经网络对体液系统的反作用,以闭环方式实现了一个含体液因子的神经网络系统;此闭环方式可以对神经网络模型的误差进行自动补偿,属于自适应神经网络的范畴。并以麻醉剂模式对这种自适应神经网络的成立条件进行推导,对其稳定性进行了验证。最后对这些含体液因子的神经网络的阈值参数进行灰延拓,拓展其成其为含灰色体液因子的神经网络系统;并把这种含灰色体液因子的神经网络系统在大规模定制生产过程的质量控制中进行了应用。

There exist three regulation methods in human organism, auto-regulation, nervous regulation and humoral regulation. This article integrates the humoral regulation to the nervous regulation and carries out the simulation of the neural network affected by humor factor. In the open loop mode, it is deduced a very useful learning formula of the threshold affected by humor system in BP neural network. Considering the back action of the neural network to the humor system, it is proposed in the paper a close loop mode system, which can eliminate the error of the neural network. In the anesthetic mode, the conditions of the loop mode is deduced and the stability of it is validated. And then the paper describes a method to extend the mixed network to a grey humor factor affecting network. Finally, the mixed system is applied to the quality control in the production process of Mass Customization.