arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1709.01971hep-lat
收藏

深度学习超越Lefschetz Thimbles

Deep Learning Beyond Lefschetz Thimbles

Andrei Alexandru, Paulo Bedaque, Henry Lamm, Scott Lawrence

摘要:

用广义顶针法处理带符号的场论问题需要反复求解计算量大的全纯流方程。我们提出了一种机器学习技术来绕过这个问题。其核心思想是通过流场方程得到几个场的结构,然后训练前馈神经网络。训练后的网络定义了一个新的集成流形,减少了符号问题,可以快速采样。本文给出了在相当大的晶格上含Wilson费米子的$1+1$维Thirring模型的结果。除了在速度上的增益,参数化的集成流形,我们使用避免了“捕获”的Monte Carlo链困扰大流量计算,以前的尝试相当大的缺点。

英文摘要:

The generalized thimble method to treat field theories with sign problems requires repeatedly solving the computationally-expensive holomorphic flow equations. We present a machine learning technique to bypass this problem. The central idea is to obtain a few field configurations via the flow equations to train a feed-forward neural network. The trained network defines a new manifold of integration which reduces the sign problem and can be rapidly sampled. We present results for the $1+1$ dimensional Thirring model with Wilson fermions on sizable lattices. In addition to the gain in speed, the parameterization of the integration manifold we use avoids the "trapping" of Monte Carlo chains which plagues large-flow calculations, a considerable shortcoming of the previous attempts.

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。

↑