论文部分内容阅读
从BP算法原理出发，找到造成这一结果的根本原因，利用目标函数对学习步长的一阶、二阶梯度值，应用牛顿近似法和线性寻优法来求得动态最优步长，这种算法所需存储一阶二阶导数的单元结构和标准BP算法中的结构相同，不会对存储造成大的负担，可使编程易于实现．计算机的仿真实验结果表明，这种方法是切实有效的．
Based on the principle of BP algorithm, find out the root cause of this result, use the objective function to calculate the first and second order gradient of the learning step, use Newton approximation and linear optimization to find the dynamic optimal step The unit structure required to store the first-order second-order derivative of the algorithm is the same as that of the standard BP algorithm, does not cause a large load on the storage, and makes the programming easy to implement. Computer simulation results show that this method is effective.