This website requires JavaScript.

ExoplANNET: A deep learning algorithm to detect and identify planetary signals in radial velocity data

L. A. NietoR. F. D\'iaz
Mar 2023
The detection of exoplanets with the radial velocity method consists indetecting variations of the stellar velocity caused by an unseen sub-stellarcompanion. Instrumental errors, irregular time sampling, and different noisesources originating in the intrinsic variability of the star can hinder theinterpretation of the data, and even lead to spurious detections. In recenttimes, work began to emerge in the field of extrasolar planets that use MachineLearning algorithms, some with results that exceed those obtained with thetraditional techniques in the field. We seek to explore the scope of the neuralnetworks in the radial velocity method, in particular for exoplanet detectionin the presence of correlated noise of stellar origin. In this work, a neuralnetwork is proposed to replace the computation of the significance of thesignal detected with the radial velocity method and to classify it as ofplanetary origin or not. The algorithm is trained using synthetic data ofsystems with and without planetary companions. We injected realistic correlatednoise in the simulations, based on previous studies of the behaviour of stellaractivity. The performance of the network is compared to the traditional methodbased on null hypothesis significance testing. The network achieves 28 % fewerfalse positives. The improvement is observed mainly in the detection ofsmall-amplitude signals associated with low-mass planets. In addition, itsexecution time is five orders of magnitude faster than the traditional method.The superior performance exhibited by the algorithm has only been tested onsimulated radial velocity data so far. Although in principle it should bestraightforward to adapt it for use in real time series, its performance has tobe tested thoroughly. Future work should permit evaluating its potential foradoption as a valuable tool for exoplanet detection.