number of additional unique samples from a distribution?

14 Views Asked by At

after drawing N times from distribution P (over a large but discrete set of things), we have K unique things.

If we draw N+M times, how many unique things will we have, in expectation?

make assumptions on the kinds of distribution of P as needed to solve the problem, for me P is a large language model, and the "unique things" are samples from the large language model.