0
$\begingroup$

Given a Young diagram $Y$, for each row $R$ choose a permutation $\sigma_R$ of $\{1,\dots, |R|\}$, where $|R|$ is the size of row $R$. Let $\sigma_R(i)$ be the “row coordinate” of the $i$th cell in row $R$.

And for each column $C$ choose a permutation $\tau_C$ of $\{1,\dots, |C|\}$, and let $\tau(j)$ be the “column coordinate” of the $j$th cell in column $C$.

The goal of choosing those permutations is to make the sum of the row coordinate and column coordinate of each cell as small as possible.

For example, when the Young diagram is a square, the sum can be as small as $|R|+1=|C|+1$ for every cell. What about other types of Young diagrams? Is there any result about this type of question?

Edit: to be more concrete, assume the diagram has rows $R_1,\dots, R_m$ (whose lengths are decreasing) and columns $C_1,\dots, C_n$ (whose lengths are decreasing). For the cell $(i,j)\in R_i\cap C_j$, let $x(i,j),y(i,j)$ be its row and column coordinates, respectively. (so $x(i,j)=\sigma_{R_i}(j)$ and $y(i,j)=\tau_{C_j}(i)$.)

Let $$t_i=\max_{(i,j)\in R_i}(x(i,j)+y(i,j)).$$

The goal is to minimize $$||t||_{\ell_1}=\sum_{i=1}^m t_i.$$

A trivial bound is $$ \sum_{i=1}^m(|R_i|+1)\le ||t|| \le \sum_{i=1}^m(|R_1|+1).$$

Any equivalent condition (or some non-trivial condition besides all the rows having the same length and $|R_1|\ge |C_1|$) that in which case, the lower bound can be attained?

$\endgroup$
6
  • $\begingroup$ When you say, "[t]he goal of choosing those permutations is to make the sum of the row coordinate and column coordinate of each cell as small as possible," I guess you mean you want to minimize the maximum over all cells of this sum of row and column coordinate? $\endgroup$ Commented Oct 28 at 20:50
  • $\begingroup$ Yes, what you said sounds more concrete: let $t$ be a vector such that $t_i$ is the maximum sum of the two coordinates of a cell in row $i$. And the target is to minimize the $\ell_1$-norm of the vector $t$. $\endgroup$
    – Connor
    Commented Oct 28 at 23:42
  • $\begingroup$ There's a trivial lower bound $\max(|R|, |C|) + 1$. It should be straightforward to adapt the argument for a square to achieve this lower bound. $\endgroup$ Commented Oct 29 at 8:19
  • $\begingroup$ Yes. Any non-trivial upper bound would be interesting. $\endgroup$
    – Connor
    Commented Oct 29 at 12:27
  • $\begingroup$ @Connor, on close reading your first comment contradicts itself and the question. If the aim is to minimise $\sum_r \max_c (\sigma_r(c) + \tau_c(r))$ then the answer to Sam's comment isn't "Yes", and the sum can't be as small as $|R| + 1$ in the case of a square, unless the square is $1 \times 1$. Please edit the question to express clearly what you actually want to ask about. $\endgroup$ Commented Nov 2 at 8:20

0

You must log in to answer this question.

Browse other questions tagged .