GSS has a huge problem set for training. The problem set consists of n types of problems, for
example dynamic programming, graph theory, data structures and so on. For the i-th type of
problem, GSS has k i problems. And each day, GSS will pick one problem from his problem set in
random and solve that. After that, he will remove that problem from the problem set and never
pick it again.
GSS found that if he has solve at least one problem from each type, he will become very moe.
(But he doesn’t know why he will become moe.) And he wants to know the expected time to solve
at least one problem from each of the n types.