Adaptive Digital Filter using NARX Deep Neural Networks for ground-based observatories

In this paper we will present a new method based on the non-symmetry of the wavelength distributions of exoplanet transits. By the clustering of wavelengths with similar behaviour, it is possible to cancel terrestrial atmospheric turbulence at ground-based observatories. Two types of benchmarks for training the NARX neural network are presented. Complexity and performance studies using transient spectroscopy data from the WASP-19b system are also included.