Quantcast
Viewing all articles
Browse latest Browse all 10

Fixed points of random permutations

The following two results are straightforward and reasonably well-known exercises in combinatorics:

  1. The number of permutations on Image may be NSFW.
    Clik here to view.
    n
    elements with no fixed points (derangements) is approximately Image may be NSFW.
    Clik here to view.
    \frac{n!}{e}
    .
  2. The expected number of fixed points of a random permutation on Image may be NSFW.
    Clik here to view.
    n
    elements is Image may be NSFW.
    Clik here to view.
    1
    .

As it turns out, it is possible to say substantially more about the distribution of fixed points of a random permutation. In fact, the following is true.

Theorem: As Image may be NSFW.
Clik here to view.
n \to \infty
, the distribution of the number of fixed points of a random permutation on Image may be NSFW.
Clik here to view.
n
elements is asymptotically Poisson with rate Image may be NSFW.
Clik here to view.
\lambda = 1
.

A boring proof

Once we know that the number of derangements of Image may be NSFW.
Clik here to view.
n
elements is approximately Image may be NSFW.
Clik here to view.
\frac{n!}{e}
, we know that the number of permutations of Image may be NSFW.
Clik here to view.
n
elements with exactly Image may be NSFW.
Clik here to view.
k
elements is Image may be NSFW.
Clik here to view.
{n \choose k}
times the number of derangements of Image may be NSFW.
Clik here to view.
n-k
elements, which is approximately

Image may be NSFW.
Clik here to view.
\displaystyle \displaystyle {n \choose k} \frac{(n-k)!}{e} = \frac{n!}{k! e}

hence the probability that a permutation on Image may be NSFW.
Clik here to view.
n
elements has Image may be NSFW.
Clik here to view.
k
fixed points is approximately Image may be NSFW.
Clik here to view.
\frac{1}{k! e}
, which is precisely the probability that a Poisson random variable with rate Image may be NSFW.
Clik here to view.
\lambda = 1
takes the value Image may be NSFW.
Clik here to view.
k
.

A more interesting proof

For this proof we will show convergence in the sense of moments; that is, if Image may be NSFW.
Clik here to view.
X_n
is the random variable describing the number of fixed points of a random permutation on Image may be NSFW.
Clik here to view.
n
elements, we will show that the moments Image may be NSFW.
Clik here to view.
\mathbb{E}(X_n^k)
converge to the moments of a Poisson random variable Image may be NSFW.
Clik here to view.
X
with rate Image may be NSFW.
Clik here to view.
\lambda = 1
. First, we compute that the moment generating function of Image may be NSFW.
Clik here to view.
X
is

Image may be NSFW.
Clik here to view.
\displaystyle \mathbb{E}(e^{tX})) = \frac{1}{e} \sum_{k \ge 0} \frac{e^{tk}}{k!} = e^{e^t - 1}

which is the exponential generating function of the Bell numbers Image may be NSFW.
Clik here to view.
B_k
, which count the number of partitions of a set with Image may be NSFW.
Clik here to view.
k
elements. We conclude that

Image may be NSFW.
Clik here to view.
\displaystyle \mathbb{E}(X^k) = B_k

so it suffices to show that Image may be NSFW.
Clik here to view.
\mathbb{E}(X_n^k)
converges to Image may be NSFW.
Clik here to view.
B_k
. Actually we will show a little more than this.

Theorem: Image may be NSFW.
Clik here to view.
\mathbb{E}(X_n^k) = B_{k, n}
is the number of partitions of a set with Image may be NSFW.
Clik here to view.
k
elements into at most Image may be NSFW.
Clik here to view.
n
non-empty subsets.

Proof 1. Note that

Image may be NSFW.
Clik here to view.
\displaystyle \mathbb{E}(X_n^k) = \frac{1}{n!} \sum_{\sigma \in S_n} \text{Fix}(\sigma)^k

where Image may be NSFW.
Clik here to view.
\text{Fix}(\sigma)^k
is the number of fixed points of Image may be NSFW.
Clik here to view.
\sigma
acting on the set of functions Image may be NSFW.
Clik here to view.
[k] \to [n]
. It follows by Burnside’s lemma that the above sum is equal to the number of orbits of Image may be NSFW.
Clik here to view.
S_n
acting on the set of functions Image may be NSFW.
Clik here to view.
[k] \to [n]
, which is precisely the number of partitions of Image may be NSFW.
Clik here to view.
[k]
into at most Image may be NSFW.
Clik here to view.
n
non-empty subsets (by taking preimages of each element of Image may be NSFW.
Clik here to view.
[n]
). Image may be NSFW.
Clik here to view.
\Box

Proof 2. By the exponential formula,

Image may be NSFW.
Clik here to view.
\displaystyle \sum_{n \ge 0} \frac{t^n}{n!} \sum_{\sigma \in S_n} z^{\text{Fix}(\sigma)} = \exp \left( z t + \frac{t^2}{2} + ... \right)

which simplifies to

Image may be NSFW.
Clik here to view.
\displaystyle \exp \left( (z - 1)t + \log \frac{1}{1 - t} \right) = \frac{1}{1 - t} \exp \left( (z - 1) t \right)
.

To extract Image may be NSFW.
Clik here to view.
k^{th}
moments, we can apply the differential operator Image may be NSFW.
Clik here to view.
\left( z \frac{\partial}{\partial z} \right)^k
and substitute Image may be NSFW.
Clik here to view.
z = 1
. Either by induction or by a combinatorial argument (see for example this math.SE question), we can show that

Image may be NSFW.
Clik here to view.
\displaystyle \left( z \frac{\partial}{\partial z} \right)^k = \sum_{r \ge 0} S(k, r) z^r \frac{\partial^r}{\partial z^r}

where Image may be NSFW.
Clik here to view.
S(k, r)
are the Stirling numbers of the second kind, which count the number of partitions of a set with Image may be NSFW.
Clik here to view.
k
elements into exactly Image may be NSFW.
Clik here to view.
r
non-empty subsets. We conclude that

Image may be NSFW.
Clik here to view.
\displaystyle \left( z \frac{\partial}{\partial z} \right)^k \frac{1}{1 - t} \exp \left( (z - 1) t \right) = \frac{1}{1 - t} \sum_{r \ge 0} S(k, r) z^r t^r \exp \left( (z - 1) t \right)

and substituting Image may be NSFW.
Clik here to view.
z = 1
we conclude that

Image may be NSFW.
Clik here to view.
\displaystyle \sum_{n \ge 0} \frac{t^n}{n!} \sum_{\sigma \in S_n} \text{Fix}(\sigma)^k = \frac{1}{1 - t} \sum_{r \ge 0} S(k, r) t^r
.

This gives

Image may be NSFW.
Clik here to view.
\displaystyle \frac{1}{n!} \sum_{\sigma \in S_n} \text{Fix}(\sigma)^k = \sum_{r \le n} S(k, r) = B_{k, n}

as desired. Image may be NSFW.
Clik here to view.
\Box


Viewing all articles
Browse latest Browse all 10

Trending Articles