We need to polish the MAB example a bit and add multilevel solutions:
[ ] Use a different response function based on the derivative of the continuous node at the 2nd level to account for the increasing uncertainty for unobserved bandits.
[ ] Add a multilevel example (as far as I can see, once #170 is merged no major technical roadblocks are on the way).
We need to polish the MAB example a bit and add multilevel solutions: