I was reading a paper (https://arxiv.org/pdf/1609.03891.pdf), and on the first page the author's write the following line: $$ \mu^\sigma = \frac{1}{n} \sum_{i = 1}^n \delta_{\left(\frac{2i}{n} - 1, \frac{2\sigma(i)}{n} -1\right)}. $$ Here, $\sigma$ is in the symmetric group on $n$ letters. The authors write that $\mu^\sigma$ is supposed to be a probability measure on $[-1, 1]^2$. I'm having trouble parsing the expression. Is the delta supposed to be the Kroenecker delta? I don't think so, because the subscript ${\left(\frac{2i}{n} - 1, \frac{2\sigma(i) }{n} -1\right)}$ doesn't seem to make sense (it would be equivalent to saying $i = \sigma(i)?$). Also, why is the domain $[-1, 1]^2$?

Could someone help me understand this expression?


Note that $$ {\left(\frac{2i}{n} - 1, \frac{2\sigma(i) }{n} -1\right)} $$ is a point in $[-1,1]^2$.
So $\delta$ with that subscript is a unit point mass at that point.
Add $n$ point masses and divide by $n$, we get a probability measure.

  • $\begingroup$ That seems to be what the diagrams in the middle of page 4 are about; they are maps of where the point masses are. $\endgroup$
    – MJD
    Dec 26 '18 at 23:58
  • $\begingroup$ @MJD So $\mu^\sigma$ places $n$ masses of $1/n$ in the square $[-1, 1]^2$ and assigns measure $0$ to every other point. Is that the right way to interpret the expression? $\endgroup$
    – Probably
    Dec 27 '18 at 0:01
  • $\begingroup$ Yes. A measure is a function on sets. The pdf is a distribution. That's the Dirac delta in $\mathbb{R}^2$ : $\delta(x,y) = \delta(x)\delta(y)$ the latter being the usual Dirac delta in $\mathbb{R}$ @Probably $\endgroup$
    – reuns
    Dec 27 '18 at 1:00

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.