Machine learning (4)---feature scaling

feature Scaling: reduces or expands the value of feature so that all feature are in a similar range, so that gradient DESCnet tend to the minimum value faster. Because the range of the different feature is very large, the contour of its cost function is elliptical, but when they are close, the contour is close to the circle, so the trend minimum is that the circle is faster, the ellipse is more tortuous, and the trend is slower. (usually within the range of +3 to-3)

Mean normalization:

Machine learning (4)---feature scaling