What approach is commonly used to extend a binary Support Vector Machine $($SVM$)$ to handle more than $2$ classes?

Group of answer choices

Using a one$-$vs$-$one or one$-$vs$-$all strategy.

Creating a separate binary SVM for each class.

Reducing the number of classes to $2$ for SVM compatibility.

Combining all classes into a single binary problem