Modeling the 3D Milky Way using Machine Learning with Gaia and infrared surveys

TL;DR

This paper presents machine learning methods to model the 3D structure of the Milky Way using Gaia and infrared surveys, improving star classification and extinction mapping.

Contribution

It introduces novel neural network-based techniques for classifying young stellar objects and reconstructing 3D extinction distribution without cross-matching datasets.

Findings

Effective YSO classification with ANN using IR data.

CNN-based 3D extinction mapping up to 10 kpc.

Potential for comprehensive Galactic plane modeling.

Abstract

The observation of our home galaxy, the Milky Way (MW), is made difficult by our internal viewpoint. The Gaia survey that contains around 1.6 billion star distances is the new flagship of MW structure and can be combined with other large-scale infrared (IR) surveys to provide unprecedented long distance measurements inside the Galactic plane. Concurrently, the past two decades have seen an explosion of the use of Machine Learning (ML) methods that are also increasingly employed in astronomy. I will first describe the construction of a ML classifier to improve a widely adopted classification scheme for Young Stellar Object (YSO) candidates. Stars being born in dense interstellar environment, the youngest ones that did not had time to move away from their formation location are a probe of the densest structures of the interstellar medium. The combination of YSO identification and Gaia distance measurements then enables the reconstruction of dense cloud structures in 3D. Our ML classifier is based on Artificial Neural Networks (ANN) and uses IR data from the Spitzer space telescope to reconstruct the YSO classification automatically from given examples. In a second part, I will propose a new method for reconstructing the 3D extinction distribution of the MW based on Convolutional Neural Networks (CNN). The CNN is trained using a large-scale Galactic model, the Besan\c{c}on Galaxy Model, and learns to infer the extinction distance distribution by comparing results of the model with observed data. This method is able to resolve distant structures up to 10 kpc with a formal resolution of 100 pc, and was found to be capable of combining 2MASS and Gaia datasets without the necessity of a cross match. The results from this combined prediction are encouraging and open the possibility for future full Galactic plane prediction using a larger combination of various datasets.