Theoretical derivation of support vector machine

Linear separable support vector machine

General process

Normalization of distance formula and introduction of extreme points

Minimization of Lagrangian dual function

KKT Conditional conclusion

SMO Algorithmic solution L Function maximization α*

Thank you, Mr. Dahai KKT deduction

• Similar to the idea of gradient descent, the optimal value is obtained through the idea of iteration , But there are only restrictions , Solve by moving two variables , More efficient .
  

Linear support vector machines

The relaxation factor is introduced and transformed into Lagrange function

KKT Conditions introduce relations and find W*

adopt KKT The relationship between b*