arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2406.02739cs.DS
收藏

具有前瞻性的局部搜索 k-means++

Local Search k-means++ with Foresight

Theo Conrads, Lukas Drexler, Joshua Könen, Daniel R. Schmidt, Melanie Schmidt

AI总结:

本文提出 Foresight LS++ 算法,通过将 LS++ 与 Lloyd 算法更好地交织,在不改变渐近运行时间和最坏情况近似界限的前提下,显著提升了 k-means 聚类在实际应用中的解质量。

摘要:

自 1957 年提出以来,用于 k-means 聚类的 Lloyd 算法已被广泛研究并经历了多项改进。虽然其原始形式完全不保证任何近似因子,但 Arthur 和 Vassilvitskii (SODA 2007) 提出了 k-means++,通过一种种子方法增强了 Lloyd 算法,在期望下保证 O(log k) 的近似。最近,Lattanzi 和 Sohler (ICML 2019) 提出了 LS++,通过局部搜索技术进一步提高了 k-means++ 的解质量,获得了 O(1) 的近似。在实践中,尽管 k-means++ 的贪婪变体在最坏情况下的表现被证明比标准 k-means++ 变体更差,但它仍常被使用。我们研究了如何在实践中进一步改进 LS++。我们研究了提高其实际性能的两种选项:(a) 将 LS++ 与贪婪 k-means++ 结合,而不是 k-means++;(b) 通过更好地将 LS++ 与 Lloyd 算法交织来改进 LS++。选项 (a) 恶化了 k-means++ 的理论保证,但如我们在实验中所证实的,它在与 LS++ 结合时也提高了实际质量。选项 (b) 是我们的新算法,Foresight LS++。我们通过实验表明,FLS++ 提高了 LS++ 的解质量。它保留了其渐近运行时间和最坏情况近似界限。

英文摘要:

Since its introduction in 1957, Lloyd's algorithm for $k$-means clustering has been extensively studied and has undergone several improvements. While in its original form it does not guarantee any approximation factor at all, Arthur and Vassilvitskii (SODA 2007) proposed $k$-means++ which enhances Lloyd's algorithm by a seeding method which guarantees a $\mathcal{O}(\log k)$-approximation in expectation. More recently, Lattanzi and Sohler (ICML 2019) proposed LS++ which further improves the solution quality of $k$-means++ by local search techniques to obtain a $\mathcal{O}(1)$-approximation. On the practical side, the greedy variant of $k$-means++ is often used although its worst-case behaviour is provably worse than for the standard $k$-means++ variant. We investigate how to improve LS++ further in practice. We study two options for improving the practical performance: (a) Combining LS++ with greedy $k$-means++ instead of $k$-means++, and (b) Improving LS++ by better entangling it with Lloyd's algorithm. Option (a) worsens the theoretical guarantees of $k$-means++ but improves the practical quality also in combination with LS++ as we confirm in our experiments. Option (b) is our new algorithm, Foresight LS++. We experimentally show that FLS++ improves upon the solution quality of LS++. It retains its asymptotic runtime and its worst-case approximation bounds.

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。

↑