[Review]SVM and related techs

Target description:

data: (x,y) Use w’x={0,1} to classify data points.

Model description

This problem can be converted to an max- margin optimization Problem:

min w’w  s.t. y(i)w’x(i)>=1, where 1/(w’w) is the margin.

Derivation: for  two points y+ and y- of class 1, 2 located on both sides of the margin boundaries.

m = 1/(2|w’w|) * (w'(y+)-w'(y-))=1/(w’w)

Leave a Reply

Your email address will not be published. Required fields are marked *