Analyzing the Kinematics of SITELLE Spectra using Machine Learning

We describe the first paper in a series of works in which we explore the application of different machine learning algorithms to the spectral analysis of extra galactic emission regions. We discuss the creation of synthetic spectra replicating the primary optical filter of the SITELLE instrument located at the Canada-France-Hawaii Telescope. We employ a convolutional neural network to learn the kinematic parameters, velocity and line broadening, directly from these spectra. Subsequently, we apply our methodology to a field of the nearby galaxy M33 and demonstrate the efficacy of our results in terms of residuals and computational expediency. We develop an open source framework for users to port this methodology to other IFUs, and we discuss future applications of machine learning to spectral analysis.