计算静电势的平衡点
Computing Equilibrium Points of Electrostatic Potentials
AI总结:
本文提出了一种基于泰勒级数分段近似的算法,用于计算静电势平衡点,证明在强非退化假设下可高效找到近似解,并探讨了该问题的复杂性分类。
摘要:
我们研究了静电势平衡点的计算问题:空间中的位置,其中由带电粒子集合产生的静电力消失。这是优化领域的一种新情况,在这种情况下,由于非构造性论证,解被保证存在,但由于奇点的存在,梯度下降法不可靠。我们提出了一种基于对势函数进行分段近似的方法,通过泰勒级数。主要思想是将域划分为具有可变粗细的网格,其中网格单元在函数变化迅速的区域比变化缓慢的区域小得多。我们的算法在近似参数的多项对数时间内找到近似的平衡点,但这些点并不保证接近精确解。然而,我们证明在一种称为“强非退化”的温和假设下,这样的点可以高效计算。我们通过研究该问题的一般化并证明其为CLS难且在PPAD中,补充了这些算法结果,留下其精确分类作为引人入胜的开放问题。
英文摘要:
We study the computation of equilibrium points of electrostatic potentials: locations in space where the electrostatic force arising from a collection of charged particles vanishes. This is a novel scenario of optimization in which solutions are guaranteed to exist due to a nonconstructive argument, but gradient descent is unreliable due to the presence of singularities. We present an algorithm based on piecewise approximation of the potential function by Taylor series. The main insight is to divide the domain into a grid with variable coarseness, where grid cells are exponentially smaller in regions where the function changes rapidly compared to regions where it changes slowly. Our algorithm finds approximate equilibrium points in time poly-logarithmic in the approximation parameter, but these points are not guaranteed to be close to exact solutions. Nevertheless, we show that such points can be computed efficiently under a mild assumption that we call "strong non-degeneracy". We complement these algorithmic results by studying a generalization of this problem and showing that it is CLS-hard and in PPAD, leaving its precise classification as an intriguing open problem.
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。