While dealing with a problem, I uncovered this subproblem:
Input: A set of sets $S = \{S_1,...,S_r\}$ where $\mid$ $S_1$ $\cup$ ... $\cup$ $S_r$$\mid = n$, as well as a number $k<n$.
Output: A subset of $S$, $T = \{S_{t_{1}},...,S_{t_{p}}\}$ so that $\mid$ $S_{t_{1}}$ $\cap$ ... $\cap$ $S_{t_{p}}$$\mid$ $\geq k$ with $p=\mid T \mid$ maximized.
I.e. find the greatest number of sets in $S$ whose intersection preserves at least $k$ elements.
Is there any way of solving this problem in reasonable time, say polynomial in $r$ and $n$?
I tried to create an algorithm to solve the problem in reasonable time for large sets and $n$ and got nowhere. There's the obvious brute force algorithm that runs in $n2^r$ time. I also tried to construct a DP algorithm that used smaller subsets for optimal substructure, i.e. making use of $(A \cap B) \cap C = A \cap B \cap C$ and that if $\mid A \cap B \mid < k$, then it's not worth searching any further with $A\cap B$. However, I didn't get anywhere since the number of different subsets grows too quickly.
On the one hand, the problem feels kind of "set coverish" and thus I suspect it's hard. On the other, I'd imagine a problem this simple to be mentioned in some list of hard problems if it were intractable.