Consider the SVM optimization problem where we explicitly allow a bias variable b$.$

In other words, the primal SVM optimization problem is stated as follows:

J $=$

n

i$=1$

max$\{0,(1$ yi$[$W $$ X

T

i $]+$ b$)\}+$

$\backslash$lambda

$2$

W$2$

Compute the dual of this optimization formulation by using analogous steps to those

discussed in the chapter. How would you handle the additional constraint in the dual

formulation during gradient descent?Consider the SVM optimization problem where we explicitly allow a bias variable $b.$

In other words, the primal SVM optimization problem is stated as follows:

$J={\displaystyle \underset{i=1}{\overset{n}{}}}$max$\{0,(1-y_i[(\frac{?}{\text{b}\text{a}\text{r}}(W))*{\stackrel{}{x}}_i^T]+b)\}+\frac{}{2}|\frac{|}{b}ar(W)|{|}^2$

Compute the dual of this optimization formulation by using analogous steps to those

discussed in the chapter. How would you handle the additional constraint in the dual

formulation during gradient descent?