$m=1$: Denote by $x_i$ the single element of $S_i$ (i.e., $S_i = \{ x_i \}$) and $X = \{ x_1, x_2, \ldots, x_L \} = \bigcup S_i$. $\displaystyle \mathbb{E}[ \max_{1\leq \ell \leq L} |S \cap S_\ell| ] = \sum_{i=1}^L (-1)^{(i-1)} sum_{S \sum_{T cap X \subseteq X_L; |T| neq \emptyset} 1 \cdot f(S) = i1 - \sum_{S \subseteq \{ 1, 2, \ldots, n \} \left(\prod_{x setminus X} f(S) = 1 - \prod_{i \in TX} x \right)$ (inclusion exclusion). (Does this necessarily lead to 1-p_i)$, which is maximized by the greedy solution of $S_i = \{ i \}$ (for $1 \leq i \leq L$) to maximize $\displaystyle \mathbb{E}[ \max_{1\leq \ell \leq L} |S \cap S_\ell| ]$?)L$).
|
6 | simplified (& furthered) analysis of 1 special case | ||
|
|
||||
|
5 | fixed typos | ||
|
$L=1$: $S_1 = \{ p_11, p_22, \ldots, p_m m \}$ maximizes $\displaystyle \mathbb{E}[ \max_{1\leq \ell \leq L} |S \cap S_\ell| ]$. $m=1$: Denote by $x_i$ the single element of $S_i$ (i.e., $S_i = \{ x_i \}$) and $X = \{ x_1, x_2, \ldots, x_L \} = \bigcup S_i$. $\displaystyle \mathbb{E}[ \max_{1\leq \ell \leq L} |S \cap S_\ell| ] = \sum_{i=1}^L (-1)^{(i-1)} \sum_{T \subseteq X_L; |T| = i} \left(\prod_{x \in T} x \right)$ (inclusion exclusion). (Does this necessarily lead to the greedy solution of $S_i = \{ p_i i \}$ (for $1 \leq i \leq L$) to maximize $\displaystyle \mathbb{E}[ \max_{1\leq \ell \leq L} |S \cap S_\ell| ]$?) Here's how I examined a few more special cases: This example ($(n,m,L)=(4,1,2)$) falls under the already discussed $m=1$ case, but I present it to hopefully make the presentation of subsequent examples clearer: This example ($(n,m,L)=(4,1,2)$) falls under the already discussed $m=1$ case, but I present it to hopefully make the presentation of subsequent examples clearer: Notes: the coefficient of 2 in the degree-4 terms of the last three rows seems interesting to me. The sign of a term is the negation of $(-1)$ to the parity of the degree of the term. |
||||
|
4 | added special cases | ||
|
EDIT: WLOG, $p_1 \geq p_2 \geq \cdots \geq p_n$. Special cases: $L=1$: $S_1 = \{ p_1, p_2, \ldots, p_m \}$ maximizes $\displaystyle \mathbb{E}[ \max_{1\leq \ell \leq L} |S \cap S_\ell| ]$. $m=1$: Denote by $x_i$ the single element of $S_i$ (i.e., $S_i = \{ x_i \}$) and $X = \{ x_1, x_2, \ldots, x_L \} = \bigcup S_i$. $\displaystyle \mathbb{E}[ \max_{1\leq \ell \leq L} |S \cap S_\ell| ] = \sum_{i=1}^L (-1)^{(i-1)} \sum_{T \subseteq X_L; |T| = i} \left(\prod_{x \in T} x \right)$ (inclusion exclusion). (Does this necessarily lead to the greedy solution of $S_i = \{ p_i \}$ (for $1 \leq i \leq L$) to maximize $\displaystyle \mathbb{E}[ \max_{1\leq \ell \leq L} |S \cap S_\ell| ]$?) Here's how I examined a few more special cases: $\binom{\binom{n}{m}}{L} = \binom{\binom{4}{1}}{2} = 6$ possibilities for $(S_1,S_2)$: |
||||
|
3 | fixed typo | ||
|
2 | fixed typo | ||
|
1 |
|
||

