用C++实现了梯度下降求多元函数极值的算法,有可能会陷入局部最优解-By c++ implementation of a gradient descent algorithm of multivariate function extreme value, is likely to fall into local optimal solution

自然梯度算法比随机梯度算法有更好的收敛性能和数值稳定性，块递归算法需要较少的运算时间。(The natural gradient algorithm works more efficiently than the ordinary gradient algorithm, and the block recursive method is feasible to real-time processing.)

基于matlab的梯度算法，直接用算子计算梯度的角度(Gradient algorithm based on MATLAB)