用于电磁兼容预测的函数链神经网络

为了解决传统BP神经网络的电磁兼容性预测方法存在易于局部收敛的问题，提出了一种基于模糊测度的函数链神经网络。该网络通过函数链将初始权重扩展到更高维度上，在实现传统BP网络多层感知的功能同时，计算过程仅为单层运算，因以收敛速度比多层的BP神经网络更快，解决了网络在解决非线性问题时，收敛于局部最小的问题。实验结果表明，提出的基于模糊测度的函数链神经网络在预测电磁兼容性方面更加精确。

Functional link neural network applied to electromagnetic compatibility prediction

Traditional prediction methods for electromagnetic compatibility of BP neural networks suffer from the local convergence. Therefore, a fuzzy measure based functional link neural network is proposed in this paper. This network extend the initial weight to a higher dimension. While achieving the multi--layer perception function of traditional BP network, the compute process is a single layer operation only. The proposed network can get faster convergence rate and avoid the conver- gence to a local minimum. The experiment result shows that the functional link neural network based on fuzzy measure can predict the electromagnetic compatibility more accurately.