Consider the training examples shown in the following table for binary classification problem:

$\backslash$table$[[$Instance$,a_1,a_2,$Target classes$],[1,T,T,+],[2,T,T,+],[3,T,F,-],[4,F,F,+],[5,F,T,-],[6,F,T,-],[7,F,F,-],[8,T,F,+],[9,F,T,-]]$

What is the entropy of this collection of training examples with respect to the positive class?

What are the information gains of $a_1$ and $a_2$ relative to these training examples?

What is the best split $($among $a_1,$ and $a_2)$ according to the information gain?