Automated tuning of RFI identification and flagging algorithms

It is known that Radio Frequency Interference affects a significant fraction of data from radio telescopes. These bad data must be carefully identified and eliminated before further processing. Current approaches require manual labor for tuning the parameters of auto flag algorithms or even marking individual regions of the data on a visual display. This paper presents a technique that can help automate the process of tuning auto flag parameters using an approach based on Artificial Intelligence, specifically on Evolutionary Computing. We describe a Genetic Algorithm that simulates the tuning of parameters throughout several generations, where no input is necessary and it produces a set of parameters that can be used in conjunction with the existing algorithms to flag the data. Experiment results demonstrate that we were able to successfully tune two auto flag algorithms for some low-frequency VLA (Very Large Array) datasets containing types of RFI that would otherwise have required manual tuning.