论文部分内容阅读
提出了一种基于预报模型的变目标函数的前向神经网络快速学习算法.首先推导出一种综合目标函数从而实现极值点附近收敛速度的提高,然后导出变目标函数的局部化递推最小二乘算法,该方法与现有同类的算法相比,可以提高收敛速度和预报精度,适用于需快速学习的系统辩识和其他应用.系统辩识的仿真实例表明了算法的优良性能“,”In this paper, we propose a fast learning algorithm of feedforward networks based on the variable criterion of predictive model. During the learning procedure, the convergence speed at the limiting value points is increased by means of the generalized cost criterion derived in part 3. The simulation results show that the predictive precision and the reliability of the model are enhanced.