Looking at this answer by Henry birthday problem - expected number of collisions and struggling to figure out why it matches this other formula provided to me on a programming related question. Thanks!
Asked
Active
Viewed 150 times
2
-
7$(1 - 1/p)^n = [(1 - 1/p)^{p}]^{n/p}$, and $(1 - 1/p)^p$ approaches $e^{-1}$ for large $p$. – Christopher A. Wong Jun 11 '14 at 01:22
-
1@ChristopherA.Wong You should make that an answer! – Jun 11 '14 at 01:48
-
you may get confused if you have seen other definitions of $e$, as maybe convergence series. There many ways to get $e$ one of the most common is the one mentioned by Christopher. – clark Jun 11 '14 at 01:54
2 Answers
3
What you may not know is that $e$ can be defined as $$\lim_{x \to \infty} (1+1/x)^x$$ from which it follows (by a little bit of limit manipulation) that $$\lim_{x \to \infty} (1+r/x)^x=e^r$$ The formula you are asking about can be explained from this; see Christopher Wong's comment above.
1
I don't know if you are familiar with Taylor series, but if yes : $$\left(1-\frac1p\right)^n=\exp\left(n\ln\left(1-\frac1p\right)\right)=\exp\left(-n\left(\frac1p+o\left(\frac1p\right)\right)\right)$$ and this gives : $$\left(1-\frac1p\right)^n\sim_{p>>1} \exp\left(-\frac{n}{p}\right)$$
Note than there's no assumption on n.
Fabien
- 2,988