In a linear combinatorial optimization problem, we are given a family $\mathcal{F} \subseteq 2^U$ of feasible subsets of a ground set $U$ of $n$ elements, and aim to find $S^* = \arg\min_{S \in \mathcal{F}} \langle w, \mathbbm{1}_S \rangle$. Traditionally, the weight vector is given, or a value oracle allows evaluating $w(S) := \langle w, \mathbbm{1}_S \rangle$. Motivated by practical interest in pairwise comparisons, and by the theoretical quest to understand computational models, we study a weaker, more robust comparison oracle that for any $S, T \in \mathcal{F}$ reveals only whether $w(S) <, =, > w(T)$. We ask: when can we find $S^*$ using few comparison queries, and when can this be done efficiently? We present three contributions: (1) We establish that the query complexity over any set system $\mathcal{F} \subseteq 2^U$ is $\tilde O(n^2)$, using the inference dimension framework, highlighting a separation between information and computational complexity (runtime may still be exponential for NP-hard problems under ETH). (2) We introduce a Global Subspace Learning (GSL) framework for objective functions with discrete integer weights bounded by $B$, giving an algorithm to sort all feasible sets using $O(nB \log(nB))$ queries, improving the $\tilde O(n^2)$ bound when $B = o(n)$. For linear matroids, algebraic techniques yield efficient algorithms for problems including $k$-SUM, SUBSET-SUM, and $A{+}B$ sorting. (3) We give the first polynomial-time, low-query algorithms for classic combinatorial problems: minimum cuts, minimum weight spanning trees (and matroid bases), bipartite matching (and matroid intersection), and shortest $s$-$t$ paths. Our work provides the first general query complexity bounds and efficient algorithms for this model, opening new directions for comparison-based optimization.
翻译:在线性组合优化问题中,给定一个由基础集$U$(包含$n$个元素)的可行子集构成的族$\mathcal{F} \subseteq 2^U$,目标是找到$S^* = \arg\min_{S \in \mathcal{F}} \langle w, \mathbbm{1}_S \rangle$。传统方法中,权重向量$w$是已知的,或者通过值预言机可计算$w(S) := \langle w, \mathbbm{1}_S \rangle$。受实际应用中成对比较需求的启发,以及探索计算模型的理论目标驱动,我们研究一种更弱、更具鲁棒性的比较预言机:对于任意$S, T \in \mathcal{F}$,该预言机仅揭示$w(S) <, =, > w(T)$的关系。我们探讨:在何种情况下能用较少的比较查询找到$S^*$,且该过程能否高效实现?本文提出三项贡献:(1)基于推理维度框架,我们证明对于任意集合系统$\mathcal{F} \subseteq 2^U$,查询复杂度为$\tilde O(n^2)$,这揭示了信息复杂度与计算复杂度之间的分离(在指数时间假设下,NP难问题的运行时间仍可能是指数级)。(2)针对离散整数权重有界(上界为$B$)的目标函数,我们提出全局子空间学习框架,给出一种使用$O(nB \log(nB))$次查询对所有可行集排序的算法,当$B = o(n)$时改进了$\tilde O(n^2)$的界。对于线性拟阵,代数技术可为包括$k$-SUM、子集和问题及$A{+}B$排序等问题提供高效算法。(3)我们首次为经典组合优化问题提出多项式时间、低查询复杂度的算法,涵盖最小割、最小权重生成树(及拟阵基)、二分图匹配(及拟阵交)以及最短$s$-$t$路径问题。本研究为该模型提供了首个通用查询复杂度界与高效算法,为基于比较的优化开辟了新方向。