Redundant interferometric calibration as a complex optimization problem
Redundant interferometric calibration as a complex optimization problem
复制标题
DOI：
10.1093/mnras/sty357
复制
发表时间：
2018-01
影响因子：
4.8
通讯作者：
T. Grobler;T. Grobler;G. Bernardi;G. Bernardi;J. Kenyon;A. Parsons;O. Smirnov
中科院分区：
文献类型：
--
作者：
T. Grobler;T. Grobler;G. Bernardi;G. Bernardi;J. Kenyon;A. Parsons;O. Smirnov
从再电离时期开始的21厘米线的红移观测最近激发了具有高度冗余配置的低频无线电阵列的构建。这些配置提供了另一种校准策略——“冗余校准”——并提高了在特定空间尺度上的灵敏度。本文将冗余干涉阵列的标定问题视为一个复杂的优化问题。我们通过Levenberg-Marquardt算法来解决这个优化问题。这种校准方法对初始条件的鲁棒性比目前的算法更强，并且通过利用近似矩阵反演，允许进一步优化和有效实现（“冗余StEfCal”）。我们还研究了使用预条件共轭梯度法作为近似矩阵逆的替代方法，但发现其计算性能与“冗余StEfCal”相比没有竞争力。这个新算法的有效实现是公开的。
Observations of the redshifted 21-cm line from the epoch of reionization have recently motivated the construction of low frequency radio arrays with highly redundant configurations. These configurations provide an alternative calibration strategy - "redundant calibration" - and boosts sensitivity on specific spatial scales. In this paper, we formulate calibration of redundant interferometric arrays as a complex optimization problem. We solve this optimization problem via the Levenberg-Marquardt algorithm. This calibration approach is more robust to initial conditions than current algorithms and, by leveraging an approximate matrix inversion, allows for further optimization and an efficient implementation ("redundant StEfCal"). We also investigated using the preconditioned conjugate gradient method as an alternative to the approximate matrix inverse, but found that its computational performance is not competitive with respect to "redundant StEfCal". The efficient implementation of this new algorithm is made publicly available.