Algorithm theory--linear SVM

Problem Introduction

The following three super-plane all play the effect of classification, which is the best one?

The answer is obviously the third one. Why?

Intuitively, if we now have a test point that is very close to the Red fork in the lower right corner, which means that the characteristic of this point is very close to that of the Red fork, then we want our classifier to divide the test point into the same class as the Red Fork.

In other words, we hope that the found hyper-plane will be able to move away from all points, that is, to minimize the distance from the super-plane to the nearest point.

We define: to any of the W,

Ask out Wi to make F (WI) equal to

How to find the maximum value

Algorithm theory--linear SVM