通过加速梯度下降法高效模拟Hartree--Fock方程
Efficient simulations of Hartree--Fock equations by an accelerated gradient descent method
AI总结:
提出三种加速策略优化梯度下降法求解Hartree--Fock方程,在氦原子模型中实现至少两个数量级加速,并优于Anderson加速方法。
摘要:
我们开发了收敛加速程序,使得梯度下降型迭代方法能够高效模拟原子之间以及原子与外部势场相互作用的Hartree--Fock方程。我们的工作聚焦于三个方面:(i)优化预处理算子中的参数;(ii)将消除最慢衰减模式的技术推广到多方程(描述多个原子)的情形;(iii)对上述技术进行新颖扩展,使得能够同时消除多个模式。我们以石墨烯片上方第一层氦原子的二维模型为例,展示了该数值方法的性能。我们证明,将上述方面(i)和(ii)纳入“朴素”梯度下降法后,其加速效果至少提高两个数量级,且通常远不止于此。方面(iii)——多模式消除——与方面(ii)的单模式消除相比,可进一步改善收敛速率。单模式和多模式消除技术均被证明显著优于著名的Anderson加速方法。我们相信,我们的加速技术也可为其他数值方法带来增益,尤其是那些处理硬核型相互作用势的方法。
英文摘要:
We develop convergence acceleration procedures that enable a gradient descent-type iteration method to efficiently simulate Hartree--Fock equations for atoms interacting both with each other and with an external potential. Our development focuses on three aspects: (i) optimization of a parameter in the preconditioning operator; (ii) adoption of a technique that eliminates the slowest-decaying mode to the case of many equations (describing many atoms); and (iii) a novel extension of the above technique that allows one to eliminate multiple modes simultaneously. We illustrate performance of the numerical method for the 2D model of the first layer of helium atoms above a graphene sheet. We demonstrate that incorporation of aspects (i) and (ii) above into the ``plain" gradient descent method accelerates it by at least two orders of magnitude, and often by much more. Aspect (iii) -- a multiple-mode elimination -- may bring further improvement to the convergence rate compared to aspect (ii), the single-mode elimination. Both single- and multiple-mode elimination techniques are shown to significantly outperform the well-known Anderson Acceleration. We believe that our acceleration techniques can also be gainfully employed by other numerical methods, especially those handling hard-core-type interaction potentials.
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。