The classic knapsack problem is maximize $P^T X$ subject to $W^T X\le M$ for $P, W\in \mathbb{R}^d$ and $M\in \mathbb{R}$ and $X\in \{0, 1\}^d$. Is there any research into algorithms that find the top $X_1, X_2, \cdots X_N$ such that $P_T X_1$, $P^T X_2, \cdots, P^T X_N$ are the top $N$ possible knapsack values subject to $W^T X_i\le M$ over all possible $X_i\in \{0, 1\}^d$? $X_1, X_2, \cdots , X_N$ need not necessarily be disjoint.
So when $W\in\mathbb{N}^d$, we have an efficient dynamic programming solution to the classic knapsack problem. Is there a similar dynamic programming solution when trying to determine the top $N$ knapsacks?
