Supervised and Semi-Supervised Methods for Exoplanet Atmospheric Parameter Retrievals from Transmission Spectroscopy

Standard bayesian retrievals for exoplanet atmospheric parameters from transmission spectroscopy, while accurate, are generally very computationally expensive. This is especially true for high-resolution spectra. There is a need for more efficient alternatives, and by leveraging modern machine learning techniques, one can optimize the retrieval process. We compare the performance of various supervised and semi-supervised machine learning methods for retrieving exoplanet atmospheric parameters from their transmission spectra. We benchmark the performance of the different algorithms on the accuracy, precision, efficiency, and computational cost. Through the use of various hyperparameters, algorithms, initializations, etc., we can quantify the uncertainties arising from the machine learning models.