RotNet: Fast and Scalable Estimation of Stellar Rotation Periods Using\n Convolutional Neural Networks

Le résumé fourni par la source

Magnetic activity in stars manifests as dark spots on their surfaces that\nmodulate the brightness observed by telescopes. These light curves contain\nimportant information on stellar rotation. However, the accurate estimation of\nrotation periods is computationally expensive due to scarce ground truth\ninformation, noisy data, and large parameter spaces that lead to degenerate\nsolutions. We harness the power of deep learning and successfully apply\nConvolutional Neural Networks to regress stellar rotation periods from Kepler\nlight curves. Geometry-preserving time-series to image transformations of the\nlight curves serve as inputs to a ResNet-18 based architecture which is trained\nthrough transfer learning. The McQuillan catalog of published rotation periods\nis used as ansatz to groundtruth. We benchmark the performance of our method\nagainst a random forest regressor, a 1D CNN, and the Auto-Correlation Function\n(ACF) - the current standard to estimate rotation periods. Despite limiting our\ninput to fewer data points (1k), our model yields more accurate results and\nruns 350 times faster than ACF runs on the same number of data points and\n10,000 times faster than ACF runs on 65k data points. With only minimal feature\nengineering our approach has impressive accuracy, motivating the application of\ndeep learning to regress stellar parameters on an even larger scale\n

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.