[论文解读] An Improved Algorithm for Dynamic Set Cover.

Sayan Bhattacharya, Monika Henzinger+2 more|arXiv (Cornell University)|Feb 25, 2020

Complexity and Algorithms in Graphs被引用 5

一句话总结

该论文提出了一种确定性的动态最小集合覆盖算法，其近似比为$(1+\epsilon)f$，更新时间为$O\left((f^2/\epsilon^3) + (f/\epsilon^2) \log C\right)$，且与全集大小$m$和集合数量$n$无关。该方法利用一种改进的动态原始-对偶框架，结合了动态顶点覆盖和集合覆盖算法的技术，实现了接近最优的更新效率。

ABSTRACT

We consider the minimum set cover problem in a dynamic setting. Here, we are given a universe of elements $\mathcal{U}$ and a collection of sets $\mathcal{S} \subseteq 2^{\mathcal{U}}$ as input, where each set $s \in \mathcal{S}$ has a cost $1/C 1$. The input keeps changing via a sequence of updates. Each update inserts/deletes an element in the universe, and specifies the sets in $§$ containing that element. The goal is to maintain a set cover of (approximately) minimum cost with as small update time as possible. Let $n$ denote the number of sets in $\mathcal{S}$, and let $m$ and $f$ respectively denote the maximum number of elements in the universe and the maximum frequency of an element across all the updates. This problem has received significant attention in recent years in the dynamic algorithms community. Gupta et al. [STOC `17] designed a deterministic $\Theta(\log m)$-approximation algorithm for this problem that has an update time of $O(f \log m)$. On the other hand, the line of work by Bhattacharya et al. [ICALP `15], Bhattacharya et al. [IPCO `17], Gupta et al. [STOC `17], Abboud et al. [STOC `19] and Bhattacharya et al. [FOCS `19] very recently led to a deterministic $(1+\epsilon)f$-approximation algorithm with $O(f \log (Cm)/\epsilon^2)$ update time. In this paper, we obtain the first dynamic algorithm for this problem with near-optimal approximation ratio whose update time is independent of $m, n$. Specifically, our algorithm is deterministic and it achieves an approximation ratio of $(1+\epsilon)f$ and an update time of $O\left((f^2/\epsilon^3)+(f/\epsilon^2) \log C ight)$. Our algorithm is based on the dynamic primal-dual framework, and it carefully combines the ideas from the dynamic vertex cover algorithm of Bhattacharya and Kulkarni [SODA `19] and the dynamic set cover algorithm of Bhattacharya et al. [FOCS `19].

研究动机与目标

• 设计一种动态集合覆盖算法，使其近似比接近理论最优值，具体为$(1+\epsilon)f$，同时最小化更新时间。
• 消除更新时间对全集大小$m$和集合数量$n$的依赖，实现接近最优的效率。
• 改进先前的动态算法，这些算法要么近似比更高，要么更新时间随$m$或$n$增长。
• 开发一种确定性算法，在元素插入和删除时能够高效维护低代价的集合覆盖。

提出的方法

• 该算法采用动态原始-对偶框架，以维护对偶变量，并在插入和删除操作中高效更新它们。
• 它整合了Bhattacharya和Kulkarni的动态顶点覆盖算法中的技术，以稳定且高效地管理对偶变量的调整。
• 该方法对Bhattacharya等人（FOCS '19）的动态集合覆盖框架进行了改进，以减少更新时间对$m$和$n$的依赖。
• 它使用一种计费机制，确保仅在必要时才更新对偶变量，从而最小化每次更新的操作次数。
• 该算法仔细平衡对偶变量的增长与原始解的维护，以实现期望的近似比。
• 它维护一个对偶解，该解是最优集合覆盖对偶解的$(1+\epsilon)$-近似，从而确保原始解的质量。

实验结果

研究问题

• RQ1 能否设计一种确定性的动态集合覆盖算法，实现$(1+\epsilon)f$的近似比，且更新时间与$m$和$n$无关？
• RQ2 如何改进动态原始-对偶框架，以消除对$m$的依赖，同时保持较低的更新成本？
• RQ3 对现有动态顶点覆盖和集合覆盖技术进行哪些修改，可实现动态集合覆盖中更高的更新效率？
• RQ4 是否可能在仅$O(f^2/\epsilon^3 + f/\epsilon^2 \log C)$的更新时间内，维持接近最优的集合覆盖？

主要发现

• 该算法实现了$(1+\epsilon)f$-近似比，与当前已知动态集合覆盖的最佳近似保证一致。
• 更新时间为$O\left((f^2/\epsilon^3) + (f/\epsilon^2) \log C\right)$，且与$m$和$n$无关，相比先前方法有显著改进。
• 该算法是确定性的，避免了可能在最坏情况更新中导致性能不可预测的随机化。
• 更新时间不随全集大小$m$或集合数量$n$增长，因此适用于大规模动态场景。
• 该方法成功结合了来自顶点覆盖和集合覆盖的动态原始-对偶技术，实现了低近似比与高效更新时间的双重优势。