arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2209.12467math.OCcs.NE
收藏

(1+1)-进化策略在梯度具有 Lipschitz 连续性的局部强凸函数上的收敛速率

Convergence rate of the (1+1)-evolution strategy on locally strongly convex functions with lipschitz continuous gradient

Daiki Morinaga, Kazuto Fukuchi, Jun Sakuma, Youhei Akimoto

AI总结:

本文分析 (1+1)-ES 在局部 L-强凸且梯度 U-Lipschitz 连续函数上的线性收敛速率,给出上下界,并证明无需目标函数先验性质。

摘要:

进化策略(ES)是用于黑盒连续优化的一类有前景的算法。尽管其在应用中取得了广泛成功,但关于其收敛速度的理论分析仍局限于凸二次函数及其单调变换。在本研究中,推导了 (1+1)-ES 在具有 U-Lipschitz 连续梯度的局部 L-强凸函数上的线性收敛速率上界和下界,分别为 exp(-Ω_{d→∞}(L/(d·U))) 和 exp(-1/d)。值得注意的是,算法不会获得关于目标函数数学性质的任何先验知识,例如 Lipschitz 常数,而现有对无导数优化算法的分析却需要这些信息。

英文摘要:

Evolution strategy (ES) is one of the promising classes of algorithms for black-box continuous optimization. Despite its broad successes in applications, theoretical analysis on the speed of its convergence is limited on convex quadratic functions and their monotonic transformation. In this study, an upper bound and a lower bound of the rate of linear convergence of the (1+1)-ES on locally $L$-strongly convex functions with $U$-Lipschitz continuous gradient are derived as $\exp\left(-Ω_{d\to\infty}\left(\frac{L}{d\cdot U}\right)\right)$ and $\exp\left(-\frac1d\right)$, respectively. Notably, any prior knowledge on the mathematical properties of the objective function, such as Lipschitz constant, is not given to the algorithm, whereas the existing analyses of derivative-free optimization algorithms require it.

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。

↑