SNIascore: Deep-learning Classification of Low-resolution Supernova Spectra

Abstract

We present SNIascore, a deep-learning based method for spectroscopic classification of thermonuclear supernovae (SNe Ia) based on very low-resolution (R∼ 100) data. The goal of SNIascore is fully automated classification of SNe Ia with a very low false-positive rate (FPR) so that human intervention can be greatly reduced in large-scale SN classification efforts, such as that undertaken by the public Zwicky Transient Facility (ZTF) Bright Transient Survey (BTS). We utilize a recurrent neural network (RNN) architecture with a combination of bidirectional long short-term memory and gated recurrent unit layers. SNIascore achieves a < 0.6% FPR while classifying up to 90% of the low-resolution SN Ia spectra obtained by the BTS. SNIascore simultaneously performs binary classification and predicts the redshifts of secure SNe Ia via regression (with a typical uncertainty of < 0.005 in the range from z = 0.01 to z = 0.12). For the magnitude-limited ZTF BTS survey (≈ 70% SNe Ia), deploying SNIascore reduces the amount of spectra in need of human classification or confirmation by ≈ 60%. Furthermore, SNIascore allows SN Ia classifications to be automatically announced in real-time to the public immediately following a finished observation during the night.

Luminous Supernovae: Unveiling a Population Between Superluminous and Normal Core-collapse Supernovae

“…We include photometry from Lunnan et al (2020) and additional photometry from PS1, which we correct for extinction by the same value of E(B −V ) = 0.009. Additional spectra from (Fremling et al 2019e) are consistent with SNe Ic. The light curve is as broad as normal SLSN-I, but much dimmer, maybe a consequence of the strong reddening.…”

Characterization of Supernovae Based on the Spectral-Temporal Energy Distribution: Possible two SN Ib Subtypes

“…A few examples for uses in the context of SN classification are Sasdelli et al (2014Sasdelli et al ( , 2016; Williamson et al (2019). Fremling et al (2021) presents a binary deep-classifier named SNIascore, which succeeds in classifying SNe as Type Ia with very low error rates. See also Baron (2019) for a useful review on machine learning methods in astronomy, including those used in this work.…”

New Modules for the SEDMachine to Remove Contaminations from Cosmic Rays and Non-target Light: byecr and contsep

“…There are advances in the use of deep learning techniques to determine the type of newly discovered transients (Charnock & Moss 2017;Moss 2018). Among them, few attempts use spectroscopic classification (e.g., Muthukrishna et al 2019;Fremling et al 2021). However, SNIascore (Fremling et al 2021) was designed specifically for spectroscopic classification focusing on SNe Ia with a training set using SEDM data.…”

“…The classification of SEDM data using SNID is currently annotated and reported to registered clients via the GROWTH Marshal (Kasliwal et al 2019) and Fritz (Duev et al 2019;van der Walt et al 2019). 8 Only SNe Ia classified by SNIascore (threshold of 0.9) are reported to the Transient Name Server (TNS) 9 (Fremling et al 2021). From the above efforts to improve the SEDM classification, we are looking forward to reporting all types of transients classified from SEDM data to TNS.…”