« First day of class » QuickCheck rocks my socks

The Poisson distribution and Stirling numbers

Posted on September 16, 2008
Tagged , , , , , , ,

While working on an assignment for my machine learning class, I rediscovered the fact that if X is a random variable from a Poisson distribution with parameter λ\lambda, then

E[Xn]=k=1nS(n,k)λk,\displaystyle E[X^n] = \sum_{k=1}^n S(n,k) \lambda^k,

where S(n,k)S(n,k) denotes a Stirling number of the second kind. (I actually prefer Knuth’s curly bracket notation, but I can’t seem to get it to work on this blog.) In particular, if λ=1\lambda = 1, then E[Xn]E[X^n] is the nth Bell number BnB_n, the number of ways of partitioning a set of size n into subsets!

As it turned out, this didn’t help me at all with my assignment, I just thought it was nifty.