LMS algorithm and gradient descent of Adline network

The LMS algorithm, which is the minimum mean variance, is the sum of squares and minima of errors.

Using gradient descent, the so-called gradient drop, essentially using the nature of the derivative to find the location of extreme points, the derivative in the vicinity of the side is greater than 0, one side is less than 0, that's all ...

And in this, the positive and negative of the derivative, is dependent on the error of the positive or negative to decide, too lazy to say, roughly

LMS algorithm and gradient descent of Adline network