Useful resource to understand question 12 of the midterm

[noranta4]

Look at this link for a nice playground to experimentally find the solutions to all the points of Question 12.
Change the number of neurons and activation functions and launch the training to see what happens in every configuration.

At this other link you can find a similar playground to look at the function learned by a MLP for a regression problem.