arXiv:2409.02943cs.DScs.LG2024-09
提出确定性算法,逼近率突破1-κ/e
A Note On Deterministic Submodular Maximization With Bounded Curvature
- 基于曲率约束的确定性优化方法
- 逼近比达(1-κ_f/e-ε),优于传统随机算法
- 适合需要可重复结果的场景
我们表明,近期[ Buchbinder and Feldman, FOCS'24]的突破性成果可进一步导出一种确定性$(1-κ_{f}/e-)$近似算法,用于在拟阵约束下最大化具有曲率$κ_{f}$的子模函数。该结果改进了传统随机算法的逼近性能,且保证了结果的确定性与可复现性。
原文摘要 · Abstract (English)
We show that the recent breakthrough result of [Buchbinder and Feldman, FOCS'24] could further lead to a deterministic $(1-κ_{f}/e-\varepsilon)$-approximate algorithm for maximizing a submodular function with curvature $κ_{f}$ under matroid constraint.
子模优化确定性算法曲率约束
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品;中文卡片由大模型生成,请以原文为准。