Applying Neural Networks to Inverse Laplace Transform Problems in Lattice QCD

将神经网络应用于格子 QCD 中的拉普拉斯逆变换问题

• 批准号: 2469788
• 金额: --
• 依托单位: University of Edinburgh
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2020
• 资助国家: 英国
• 起止时间: 2020 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2469788
• 关键词: Applying; Neural; Networks; Inverse; Laplace

Correlation functions obtained from Lattice QCD are Laplace transforms of corresponding spectral density functions. These spectral densities are central to understanding the underlying physics, so the problem of extracting them from Lattice QCD correlation functions, i.e. inverting these Laplace transforms, is of great interest. This is a very ill-conditioned problem however, with most methods being susceptible to large numerical errors. The aim of this project is to use the potential of neural networks and machine learning to improve the accuracy and reliability of the existing methods for inverting Laplace transforms and thus be able to extract the associated spectral density functions with lower numerical errors.

从 Lattice QCD 获得的相关函数是相应谱密度函数的拉普拉斯变换。这些谱密度对于理解基础物理至关重要，因此从格子 QCD 相关函数中提取它们（即反转这些拉普拉斯变换）的问题非常有趣。然而，这是一个非常病态的问题，大多数方法都容易受到较大的数值误差的影响。该项目的目的是利用神经网络和机器学习的潜力来提高现有拉普拉斯变换反演方法的准确性和可靠性，从而能够以较低的数值误差提取相关的谱密度函数。