Nichols Nathan S, Sokol Paul, Del Maestro Adrian
Data Science and Learning Division, Argonne National Laboratory, Argonne, Illinois 60439, USA.
Department of Physics, University of Vermont, Burlington, Vermont 05405, USA.
Phys Rev E. 2022 Aug;106(2-2):025312. doi: 10.1103/PhysRevE.106.025312.
We report on differential evolution for analytic continuation: a parameter-free evolutionary algorithm to generate the dynamic structure factor from imaginary time correlation functions. Our approach to this long-standing problem in quantum many-body physics achieves enhanced spectral fidelity while using fewer compute (CPU) hours. The need for fine-tuning of algorithmic control parameters is eliminated by embedding them within the genome to be optimized for this evolutionary computation-based algorithm. Benchmarks are presented for models where the dynamic structure factor is known exactly and experimentally relevant results are included for quantum Monte Carlo simulations of bulk ^{4}He below the superfluid transition temperature.
一种从虚时关联函数生成动态结构因子的无参数进化算法。我们针对量子多体物理中这个长期存在的问题所采用的方法，在使用更少计算（CPU）时间的同时提高了光谱保真度。通过将算法控制参数嵌入到基因组中，从而为基于进化计算的算法进行优化，消除了对算法控制参数进行微调的需求。文中给出了动态结构因子已知精确值的模型的基准测试，并包含了超流转变温度以下体相(^4)He的量子蒙特卡罗模拟的实验相关结果。
