To improve generalization, we use a tool that is called regularization.

In simple words,

$L_{\text{l}\text{o}\text{s}\text{s}}(w)=L_{logistic-\text{r}\text{e}\text{g}}(w)+*||w|{|}^2.$

Change 'Logistic$\_$Regression$\_$via$\_$GD$\text{'}$ according to that loss and find the best hyperparameter $$ using the performance on the validation.

lamads $=np.$arange$(0,5,0\mathrm{.}1)$

for lamda in lamads:

#Solve logistic regression with lamda

print $($f$"$Valdation accuracy for lamda$=\{$lamda:$\mathrm{.}2$f$\}$: $\{$accuracy $***100)$