The Impact of Host Galaxy Properties on Supernova Classification with Hierarchical Labels Astropy ( 43 ; 3 ; 4 ) , FLEET ( 19 ) , Matplotlib ( 27 ) , NumPy ( 23 ) , Superphot ( 24 ) , SuperRAENN ( 48 ) , Pytorch ( 39 )

V. Ashley Villar, Sebastian Gomez, Edo Berger, Alex Gagliano · 2026

Read The Impact of Host Galaxy Properties on Supernova Classification with Hierarchical Labels Astropy ( 43 ; 3 ; 4 ) , FLEET ( 19 ) , Matplotlib ( 27 ) , NumPy ( 23 ) , Superphot ( 24 ) , SuperRAENN ( 48 ) , Pytorch ( 39 ) by V. Ashley Villar, Sebastian Gomez, Edo Berger, Alex Gagliano (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-24807.txt.

Introduction

V. Ashley Villar, Sebastian Gomez, Edo Berger, and Alex Gagliano test whether a supernova's host galaxy helps classify the explosion from light alone. The Vera Rubin Observatory will discover supernovae far faster than spectrographs can follow them up. Photometric classifiers must therefore sort events in real time and for archives. Most use light curves, sometimes with redshift. This study asks what observable host galaxy traits add. Hierarchical class labels, respecting the family tree of supernova types, structure the task.

Host galaxy information alone isolates fairly pure samples of Type Ia supernovae, above ninety percent purity, with or without redshift. With redshift, samples of Type II and superluminous supernovae pass seventy percent purity. When full light curves are present, host traits add little accuracy if redshift is known. Without redshift, they raise classification accuracy significantly. The team also introduces a weighted hierarchical cross entropy loss for the task.

Catalog record

CREATORV. Ashley Villar, Sebastian Gomez, Edo Berger, Alex Gagliano
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE57 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.24807

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

The Impact of Host Galaxy Properties on Supernova Classification with Hierarchical Labels Astropy ( 43 ; 3 ; 4 ) , FLEET ( 19 ) , Matplotlib ( 27 ) , NumPy ( 23 ) , Superphot ( 24 ) , SuperRAENN ( 48 ) , Pytorch ( 39 ) V. Ashley Villar, Sebastian Gomez, Edo Berger, Alex Gagliano Abstract With the advent of the Vera C. Rubin Observatory, the discovery rate of supernovae (SNe) will surpass the rate of SNe with real time spectroscopic followup by three orders of magnitude. Accurate photometric classifiers are essential to both select interest events for followup in real time and for archival population-level studies. In this work, we investigate the impact of observable host galaxy information on the classification of SNe, both with and without additional light curve and redshift information. We find that host galaxy information alone can successfully isolate relatively pure ( > 90 >90 %) samples of Type Ia SNe with or without redshift information. With redshift information, we can additionally produce somewhat pure ( > 70 >70 %) samples of Type II SNe and superluminous supernovae. Additionally with redshift information, host galaxy properties do not significantly improve the accuracy of SN classification when paired with complete light curves. In the absence of redshift information, however, galaxy properties significantly increase the accuracy of photometric classification. As a part of this analysis, we present the first formal application of a new objective function, the weighted hierarchical cross-entropy, to the problem of supernova classification. This objective function more naturally accounts for the hierarchical nature of supernova classes and, more broadly, transients. Finally, we present a new set of SN classifications for the Pan-STARRS Medium Deep Survey of SNe that lack spectroscopic redshift, increasing the full photometric sample to > > 4400 events. I Introduction Early classification of extragalactic transients, in particular supernovae (SNe), is paramount to enabling multi-wavelength and spectroscopic analysis in real time. Currently, \sim 10 percent of SNe receive a spectroscopic classification, the traditional means for understanding the underlying nature of a SN. The fraction of SNe which remain spectroscopically unclassified will significantly grow with the upcoming Legacy Survey of Space and Time (LSST) conducted by Vera C. Rubin Observatory and expected to commence in 2025. LSST will discover over one million SNe annually ( 31 ) ; without additional spectroscopic resources, \lesssim 0.1\% of all SNe will have spectroscopic followup. Large-scale spectroscopic followup campaigns, such as 4MOST ( 46 ) , will increase this to \sim 1 %. As a result of this mismatch in discovery vs followup rates, the time-domain community has placed significant emphasis on photometric classifiers in the years leading up to LSST. Photometric classifiers aim to classify SNe into their historically spectroscopic classes based on photometric data alone. The taxonomy of SNe has actively evolved as increasingly large samples unveil new diversity in SNe observables, leading to a branching hierarchical structure. Broadly speaking, Type II SNe are those which show spectroscopic signatures of hydrogen near the peak of their optical emission. Within this class, the spectra of Type IIP/L SNe have broad H (typically with P-Cygni profiles), while Type IIn SNe are dominated by narrow H emission. In contrast, Type I SNe lack said features, with Type Ib SNe showing signs of helium and Type Ic SNe lacking signs of either. Type Ibc and II SNe arise from the core-collapse of massive stars. Type Ia SNe, arising from thermonuclear explosions of white dwarfs, lack H and He in their optical spectra but show strong signatures of Si II near peak. A broad overview of the hierarchical SN taxonomy is presented in 17 . Some photometric classification methods can be conducted in real time (e.g., 38 ; 37 ; 11 ; 41 ; 16 ) while others rely on complete light curves (e.g., 24 ; 9 ) for feature extraction. It can be especially challenging to identify and measure useful SN features in the early photospheric phases of the light curves. However, prompt identification of SNe is key to capture early observational phenomena (e.g., flash spectroscopy for young core-collapse SNe; 32 ; 10 ; 28 ), to guide multiwavelength follow up and to optimize spectroscopic resources. In order to perform this early classification, all available information must be utilized, including the contextual information provided by the host galaxy of the transient. Specifically, host galaxy information is known to correlate with transient properties. For example, while thermonuclear (Type Ia) SNe are broadly observed across all galaxy types, core-collapse supernovae (CCSNe) occur only in galaxies with recent or ongoing star formation ( 36 ; 22 ; 12 ) . Type Ib and Ic SNe, specific substypes of stripped-envelope SNe (SESNe), prefer slightly more massive (and typically higher metallicity) host galaxies compared to their hydrogen-rich counterparts of Type II(b) SNe ( 45 ) . Similarly, SESNe are over-represented in disturbed galaxies compared to their H-rich counterparts ( 21 ) . Rare CCSN classes seem to show stronger and more exotic preferences. Both Type Ic-BL (high-energy, “broad-lined” events) and H-poor superluminous SNe (SLSNe) prefer low mass (metal poor) galaxies with high specific star formation rates ( 30 ; 45 ) . On local scales, SESNe strongly trace H \alpha (a tracer of ongoing star formation), while Type II SNe tend to show larger spreads in local environment properties 2 . In contrast the Type IIn SNe, which show signs of interaction between the SN blast wave and dense, pre-existing circumstellar material, show highly heterogeneous local environments ( 42 ) . The rarer Ca-rich SNe of yet unknown progenitor origin show strong preference for high offsets in their host galaxies, although they are not preferentially found in star-forming hosts ( 29 ; 13 ) . These correlations are clues toward the underlying progenitor populations for each of these SN classes. Here, however, we utilize the correlations between global host properties and transients as a tool to help classify the underlying physics of the SNe. Classifiers that utilize galaxy information for SN identification have been explored in the literature, although they have been primarily limited to binary classification tasks. 14 introduced the first galaxy-based SN classifier based on a Naive Bayes architecture with the LOSS SN sample. In a similar vein, 5 showed that galaxy morphology and star formation could be used to increase the purity of thermonuclear and CCSN samples compared to randomly guessing. 15 introduced a random forest classifier based on host galaxy properties from the GHOST sample of \simeq 16,000 SNe associated with host galaxies, finding that such a method could perform thermonuclear vs CCSNe classification with \simeq 68 % accuracy. 19 presented one of the first host-aware classifiers which extend beyond Ia/CCSNe classification, focusing instead on Type I SLSNe vs non-SLSN classification from a combination of host and light curve properties; a classifier based on host galaxy properties alone is not explored in that work. 34 presented a hierarchical SN classifier based solely on optical and (near) infrared host galaxy photometry from the THEx catalog ( 40 ) . They use a likelihood-based approach reminiscent of a Naive Bayes classifier, using a series of binary classifications to distinguish 11 SN classes (in addition to tidal disruption events). They

Read the complete text