Pointing Model Meets Deep Learning: A Retrospective Study on a MeerKAT+ Telescope Applying Deep Learning Methods for Blind Pointing Corrections

Abstract

This study aims to compare the effectiveness of deep learning methods, specifically Feedforward Neural Networks (FNNs), with traditional Pointing Models (PMs) for compensating Blind Pointing Errors in astronomical instruments. Ambitious projects like the ongoing study for the Atacama Large Aperture Submillimeter Telescope (AtLAST) inspired the investigation of possible improvements to traditional Pointing Error (PE) modeling. The study assesses the practicality of FNNs by applying them to data from an instrument in operation: a precursor MeerKAT+ telescope from the Max Planck Institute for Radio Astronomy (MPIfR), intended to extend the current MeerKAT Radio Telescope Array at the South African Radio Astronomy Observatory (SARAO) site in the Meerkat National Park in South Africa.