迭代算法-梯度下降法Python理解

 

 

对theta1求偏导：

 1 def partial_cost_theta1(theta0, theta1, x, y):
 2     # Hypothesis
 3     h = theta0 + theta1*x
 4     # Hypothesis minus observed times x
 5     diff = (h - y) * x
 6     # Average to compute partial derivative
 7     partial = diff.sum() / (x.shape[0])
 8     return partial
 9 
10 =============================
11 partial = diff.sum() / (x.shape[0])
12 这是梯度下降法中用python对theta0求偏导，
13 其中，
14 a)diff=(h-y)涉及到的参数x与y均是传入的数组类型，所以在python语法中可以用diff.sum()对这些值进行求和操作；
15 b)x.shape[0]，表示求出这个列数组（矩阵）的维度数（n维1列）.

 

posted @ 2017-07-02 22:53  海之涯2008  阅读(755)  评论(0)  编辑  收藏  举报