On Feed-Forward Neural-Network-Based Atmospheric Tomography

Imaging an astronomical object from a ground location suffers from atmospheric turbulence degrading the quality of the image. Adaptive optics systems use bright guide-stars to estimate the aberration and correct for it, improving the imaging quality. For objects with an angular distance outside the isoplanatic patch of a bright guide-star, atmospheric tomography can be employed as an extension to adaptive optics, estimating the aberration for those targets. This paper explores the efficacy of a feed-forward neural-network-based model for reconstruction of the target phase aberration, showing that it performs better than a least-squares atmospheric tomography method.