梯度下降法是一个最优化算法，通常也称为最速下降法。最速下降法是求解无约束优化问题最简单和最古老的方法之一，虽然现在已经不具有实用性，但是许多有效算法都是以它为基础进行改进和修正而得到的。最速下降法是用负梯度方向为搜索方向的，最速下降法越接近目标值，步长越小，前进越慢。(The gradient descent method is an optimization algorithm, also known as steepest descent. The steepest descent method is one of the methods for solving unconstrained optimization problems is the most simple and most ancient, although it is not practical, but many effective algorithms are based on it were improved and revised by. The steepest descent method uses the negative gradient direction as the search direction, and the steepest descent method is closer to the target value, and the smaller the step length, the slower the advance.)