1
$\begingroup$

Draw $n$ numbers, denoted by $a_1, a_2, \ldots, a_n$, from set $[n]$, that is, for each $i$, $a_i$ is a uniformly random number from $[n]$.

Let $A = \{a_1, a_2, \ldots, a_n\}$. Then $$ \mathbb{E}[|A|] = n - n (\frac{n-1}{n})^n \approx n (1 - 1/e). $$ How to prove a concentration bound on $|A|$? For example, prove $\Pr[|A| < n / 100] < 2^{-\Omega(n)}$.

$\endgroup$
2
$\begingroup$

We have that $$ P(|A|\le cn) \le \binom{n}{cn} c^n $$ (first choose a subset of cardinality $cn$ and then insist that all numbers come from this set). By Stirling's formula, this bound is asymptotically $$ \sim \frac{1}{\sqrt{n}} \left( \frac{c}{1-c} \right)^{(1-c)n} , $$ which is of the requested type as long as $c<1/2$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.