arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2412.03332cs.DScs.CCcs.CGcs.LG
收藏

关于 $\ell_2^2$ 最小和聚类的可近似性

On Approximability of $\ell_2^2$ Min-Sum Clustering

Rutgers University(罗格斯大学);University of Michigan(密歇根大学);The Hebrew University of Jerusalem(耶路撒冷希伯来大学);Aarhus University(奥胡斯大学);Texas A&M University(德克萨斯农工大学)

Karthik C. S., Euiwoong Lee, Yuval Rabani, Chris Schwiegelshohn, Samson Zhou

AI总结:

本文首次证明 $\ell_2^2$ 最小和 $k$-聚类在优于 1.056(及条件性 1.327)因子下的近似是 NP-hard 的,同时给出近线性时间参数化 PTAS 和学习增强设置下的多项式时间近似算法。

摘要:

$\ell_2^2$ 最小和 $k$-聚类问题是将输入集合划分为簇 $C_1,\ldots,C_k$,以最小化 $\sum_{i=1}^k\sum_{p,q\in C_i}\\|p-q\\|_2^2$。尽管 $\ell_2^2$ 最小和 $k$-聚类是 NP-hard 的,但尚不清楚在超过某个因子后近似 $\ell_2^2$ 最小和 $k$-聚类是否 NP-hard。本文首次给出了 $\ell_2^2$ 最小和 $k$-聚类问题的近似难度结果。我们证明,将目标函数近似到优于 $1.056$ 的因子是 NP-hard 的;此外,在 Johnson Coverage Hypothesis 的平衡变体假设下,将目标函数近似到优于 1.327 的因子也是 NP-hard 的。随后,我们通过给出一个近线性时间的参数化 PTAS 来补充上述难度结果,该算法运行时间为 $O\left(n^{1+o(1)}d\cdot \exp((k\cdot\varepsilon^{-1})^{O(1)})\right)$,其中 $d$ 是输入数据集的底层维度。最后,我们考虑学习增强设置,其中算法可以访问一个预言机,该预言机为输入点输出标签 $i\in[k]$,从而隐式地将输入数据集划分为 $k$ 个簇,这些簇诱导出一个近似最优解,误差至多为某个对抗性误差 $α\in\left[0,\frac{1}{2}\right)$。我们给出一个多项式时间算法,对于固定常数 $γ>0$,输出 $\ell_2^2$ 最小和 $k$-聚类的 $\frac{1+γα}{(1-α)^2}$-近似。

英文摘要:

The $\ell_2^2$ min-sum $k$-clustering problem is to partition an input set into clusters $C_1,\ldots,C_k$ to minimize $\sum_{i=1}^k\sum_{p,q\in C_i}\|p-q\|_2^2$. Although $\ell_2^2$ min-sum $k$-clustering is NP-hard, it is not known whether it is NP-hard to approximate $\ell_2^2$ min-sum $k$-clustering beyond a certain factor. In this paper, we give the first hardness-of-approximation result for the $\ell_2^2$ min-sum $k$-clustering problem. We show that it is NP-hard to approximate the objective to a factor better than $1.056$ and moreover, assuming a balanced variant of the Johnson Coverage Hypothesis, it is NP-hard to approximate the objective to a factor better than 1.327. We then complement our hardness result by giving a nearly linear time parameterized PTAS for $\ell_2^2$ min-sum $k$-clustering running in time $O\left(n^{1+o(1)}d\cdot \exp((k\cdot\varepsilon^{-1})^{O(1)})\right)$, where $d$ is the underlying dimension of the input dataset. Finally, we consider a learning-augmented setting, where the algorithm has access to an oracle that outputs a label $i\in[k]$ for input point, thereby implicitly partitioning the input dataset into $k$ clusters that induce an approximately optimal solution, up to some amount of adversarial error $α\in\left[0,\frac{1}{2}\right)$. We give a polynomial-time algorithm that outputs a $\frac{1+γα}{(1-α)^2}$-approximation to $\ell_2^2$ min-sum $k$-clustering, for a fixed constant $γ>0$.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑