Exoplanet Detection by Machine Learning with Data Augmentation by Koray Aydoğan. on Wednesday 30 NovemberIt has recently been demonstrated that deep learning has significantpotential to automate parts of the exoplanet detection pipeline using lightcurve data from satellites such as Kepler \cite{borucki2010kepler}\cite{koch2010kepler} and NASA's Transiting Exoplanet Survey Satellite (TESS)\cite{ricker2010transiting}. Unfortunately, the smallness of the availabledatasets makes it difficult to realize the level of performance one expectsfrom powerful network architectures. In this paper, we investigate the use of data augmentation techniques onlight curve data from to train neural networks to identify exoplanets. Theaugmentation techniques used are of two classes: Simple (e.g. additive noiseaugmentation) and learning-based (e.g. first training a GAN\cite{goodfellow2020generative} to generate new examples). We demonstrate thatdata augmentation has a potential to improve model performance for theexoplanet detection problem, and recommend the use of augmentation based ongenerative models as more data becomes available.arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.15577v1