Characterization of exoplanets' light curves with neural networks

Abstract

A study of the implementation of deep learning using artificial neural networks is undertaken aiming to reduce processing time and human supervision in the characterization of exoplanets’ light curves. Firstly, to understand the problem and the techniques involved, a convolutional neural network proposed by Shallue & Vanderburg for the Kepler mission is studied and recreated. Secondly, different alternative neural networks are proposed and compared with the original one, aiming to improve the classification performance