Exploring the spectroscopic diversity of type Ia supernovae with DRACULA: a machine learning approach

The existence of multiple subclasses of type Ia supernovae (SNeIa) has been the subject of great debate in the last decade. One major challenge inevitably met when trying to infer the existence of one or more subclasses is the time consuming, and subjective, process of subclass definition. In this work, we show how machine learning tools facilitate the automatic discovery of sub-populations of SNIa; to that end we introduce the DRACULA Python package (Dimensionality Reduction And Clustering for Unsupervised Learning in Astronomy). Our approach is divided in three steps: (i) Transfer Learning, which takes advantage of all available spectra (even those without an epoch estimate) as an information source, (ii) dimensionality reduction through Deep Learning and (iii) unsupervised learning (clustering) using K-Means. Results match a previously suggested classification scheme, showing that the proposed method is able to grasp the main spectral features behind the definition of such subclasses. Moreover, our methodology is capable of automatically identifying a hierarchical structure of spectral features. This allows the confirmation of the velocity of lines as a first order effect in the determination of SNIa sub-classes, followed by 91bg-like events. In this context, SNIa spectra are described by a space of 4 dimensions + 1 for the time evolution of objects. We interpreted this as evidence that the progenitor system and the explosion mechanism should be described by a small number of initial physical parameters. Given the expected data deluge in the forthcoming years, our proposed approach is essential to allow a quick and statistically coherent identification of subclasses (and outliers). DRACULA is publicly available within COINtoolbox (https://github.com/COINtoolbox/DRACULA).