Inverse Huffman
is an exercise over the optimal encoding of a system of information, by a
code of variable lengths. But the style of the exercise is in the inverse
direction: the server gives you a code, and asks you to find a distribution
of probabilities such that the given code is optimal.