Neural network can precisely approach the function and inverse function of multivariate nonlinear equation groups,and solve the groups by using iterative algorithm. In the experiment,the endless iterations caused by premature conver⁃gence of error was found in the iteration process,and it is also found that the value of convergence not always is the minimum. The iterative algorithm is improved for the two problems,and the experimental analysis is performed by using examples. The ap⁃proximate solution of the equation can be got even if input sample interval deviates from the right solution very far.% 利用BP神经网络可以逼近多变元非线性方程组的函数的反函数,并利用迭代算法进行求解方程组的根。在实验中发现迭代算法在迭代过程中由于误差过早收敛出现无穷迭代,且由于误差并不一直随迭代次数的增加而减少导致最后收敛的值并不一定是最小值。针对这两个问题对其进行了改进,并利用算例进行了实验分析。在输入区域偏离解区域很远的情况下也能求得方程组的近似解。
展开▼