Ordered-subsets Acceleration of Radio Interferometric Calibration: Os-sage Calibration Algorithm

The main objective of this work is to accelerate the MaximumLikelihood (ML) estimation procedure for radio interferometric direction dependent self-calibration. We introduce the OS-SAGE radio interferometric calibration method, as a combination of the Ordered-Subsets (OS) method with the Space Alternating Generalized Expectation maximization (SAGE) calibration technique. The OS method speeds up the ML estimation and achieves nearly the same level of accuracy of solutions as the one obtained by the nonOS methods. However, the method usually gets stuck at a limit cycle of some suboptimal points and does not converge to a global optimum. We achieve a higher accuracy of calibration by introducing a modified OS-SAGE calibration scheme in which, at the last iteration of the calibration procedure, the suboptimal values obtained by the OS method are considered as the calibration solutions. That ends up the calibration of every dataset with as many solutions as there are sub-observations in that dataset, rather than a common solution, thus improving the calibration’s accuracy.