子集和比问题（SSR）旨在给定一个正整数多重集合 \( A \) 的情况下，找到两个不相交的子集，使得它们的和的最大到最小比值最小化。本文研究了 \( k \) 版本的 SSR，即 \( k \)-子集和比问题 (\( k \)-SSR)，其目标是最小化 \( A \) 的 \( k \) 个不相交子集的和的最大到最小比值。我们开发了一个运行时间为 \( O({n^{2k}}/{\varepsilon^{k-1}}) \) 的近似方案，其中 \( n = |A| \) 且 \( \varepsilon \) 是误差参数。据我们所知，这是固定 \( k > 2 \) 时首个针对 \( k \)-SSR 的完全多项式时间近似方案（FPTAS）。我们还提出了 \( k \)-路数划分比问题 (\( k \)-PART) 的一个 FPTAS，该问题与 \( k \)-SSR 的区别在于 \( k \) 个子集必须构成 \( A \) 的划分。我们为 \( k \)-PART 提出了一个更复杂的 FPTAS，同样达到了 \( O({n^{2k}}/{\varepsilon^{k-1}}) \) 的时间复杂度。值得注意的是，\( k \)-PART 等价于具有相同估值函数的最小嫉妒比问题，在不可分物品公平分配的研究背景下已被研究过。当限制在相同估值的情况下，我们的 FPTAS 相较于 Nguyen 和 Rothe 的针对最小嫉妒比的 FPTAS 有显著改进，后者对于所有加性估值函数的时间复杂度为 \( O(n^{4k^2+1}/\varepsilon^{2k^2}) \)。最后，我们为 \( k \)-SSR 提出了一种第二种 FPTAS，它通过精心设计调用第一种方案实现；新的方案时间复杂度为 \( \widetilde{O}(n/{\varepsilon^{3k-1}}) \)，因此当 \( n \gg 1/\varepsilon \) 时比第一种方案快得多。

The Subset Sum Ratio problem (SSR) asks, given a multiset $A$ of positive integers, to find two disjoint subsets of $A$ such that the largest-to-smallest ratio of their sums is minimized. In this paper, we study the $k$-version of SSR, namely $k$-Subset Sum Ratio ($k$-SSR), which asks to minimize the largest-to-smallest ratio of sums of $k$ disjoint subsets of $A$. We develop an approximation scheme for $k$-SSR running in $O({n^{2k}}/{\varepsilon^{k-1}})$ time, where $n=|A|$ and $\varepsilon$ is the error parameter. To the best of our knowledge, this is the first FPTAS for $k$-SSR for fixed $k>2$. We also present an FPTAS for the $k$-way Number Partitioning Ratio ($k$-PART) problem, which differs from $k$-SSR in that the $k$ subsets must constitute a partition of $A$. We present a more involved FPTAS for $k$-PART, also achieving $O({n^{2k}}/{\varepsilon^{k-1}})$ time complexity. Notably, $k$-PART is equivalent to the minimum envy-ratio problem with identical valuation functions, which has been studied in the context of fair division of indivisible goods. When restricted to the case of identical valuations, our FPTAS represents a significant improvement over Nguyen and Rothe's FPTAS for minimum envy-ratio, which runs in $O(n^{4k^2+1}/\varepsilon^{2k^2})$ time for all additive valuations. Lastly, we propose a second FPTAS for $k$-SSR, which employs carefully designed calls to the first one; the new scheme has a time complexity of $\widetilde{O}(n/{\varepsilon^{3k-1}})$, thus being much faster than the first one when $n\gg 1/ \varepsilon$.