arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1404.2787stat.APmath.STphysics.data-anstat.TH
收藏

线性迭代展开法的收敛性和误差传播

Convergence and error propagation results on a linear iterative unfolding method

Andras Laszlo

摘要:

展开问题通常出现在统计数据分析的背景下。当要测量物理量的概率分布时,会出现这样的问题,但是它被一些很好理解的过程随机化(模糊化),例如非理想的检测器响应或很好描述的物理现象。在这种情况下,可以说感兴趣的原始概率分布被已知的响应函数折叠。从测量的概率分布重建原始的概率分布称为展开。从技术上讲,这涉及到计算L^1函数空间上积分算子的无界逆,这是一个不适定问题。对于相关的正则化算子反演,我们提出了一个线性迭代公式,并在概率论的背景下提供收敛性证明。此外,我们提供了有限迭代停止阶的误差估计公式,这些公式在实际应用中至关重要:近似误差、传播的统计误差和传播的系统误差可以量化。这些论点是基于将原始L^1问题映射到L^2空间的Riesz-Thorin定理,以及随后普通L^2算子谱理论的应用。还提供了该算法的C语言库实现以及相应的错误传播。数值例子也说明了该方法的操作。

英文摘要:

Unfolding problems often arise in the context of statistical data analysis. Such problematics occur when the probability distribution of a physical quantity is to be measured, but it is randomized (smeared) by some well understood process, such as a non-ideal detector response or a well described physical phenomenon. In such case it is said that the original probability distribution of interest is folded by a known response function. The reconstruction of the original probability distribution from the measured one is called unfolding. That technically involves evaluation of the non-bounded inverse of an integral operator over the space of L^1 functions, which is known to be an ill-posed problem. For the pertinent regularized operator inversion, we propose a linear iterative formula and provide proof of convergence in a probability theory context. Furthermore, we provide formulae for error estimates at finite iteration stopping order which are of utmost importance in practical applications: the approximation error, the propagated statistical error, and the propagated systematic error can be quantified. The arguments are based on the Riesz-Thorin theorem mapping the original L^1 problem to L^2 space, and subsequent application of ordinary L^2 spectral theory of operators. A library implementation in C of the algorithm along with corresponding error propagation is also provided. A numerical example also illustrates the method in operation.

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。

↑