DeepZipper: A Novel Deep Learning Architecture for Lensed Supernovae Identification

Abstract

Large-scale astronomical surveys have the potential to capture data on large numbers of strongly gravitationally lensed supernovae (LSNe). To facilitate timely analysis and spectroscopic follow-up before the supernova fades, an LSN needs to be identified soon after it begins. To quickly identify LSNe in optical survey datasets, we designed ZipperNet, a multi-branch deep neural network that combines convolutional layers (traditionally used for images) with long short-term memory (LSTM) layers (traditionally used for time series). We tested ZipperNet on the task of classifying objects from four categories -no lens, galaxy-galaxy lens, lensed type Ia supernova, lensed core-collapse supernova -within high-fidelity simulations of three cosmic survey data sets -the Dark Energy Survey (DES), Rubin Observatory's Legacy Survey of Space and Time (LSST), and a Dark Energy Spectroscopic Instrument (DESI) imaging survey. Among our results, we find that for the LSST-like dataset, Zip-perNet classifies LSNe with a receiver operating characteristic area under the curve of 0.97, predicts the spectroscopic type of the lensed supernovae with 79% accuracy, and demonstrates similarly high performance for LSNe 1-2 epochs after first detection. We anticipate that a model like ZipperNet, which simultaneously incorporates spatial and temporal information, can play a significant role in the rapid identification of lensed transient systems in cosmic survey experiments.

Cosmology Intertwined: A Review of the Particle Physics, Astrophysics, and Cosmology Associated with the Cosmological Tensions and Anomalies

“…These are the two main observational properties which can be used to select plausible glSN candidates. Novel techniques can be employed to help identify glSNe with partially blended images, such as neural networks trained on time series of images [1943,1944]. Another strategy for finding glSNe is to monitor known lensed galaxies [1945].…”

Machine Learning and Cosmology

“…classification of light curves from Type 1a supernovae). Recurrent Neural Networks [117], Long Short-Term Memory Networks [118], and Transformers [119] have made significant advances for classifica-tion problems, but sparsely sampled light curves with noisy observations present many challenges for standard neural network architectures common in applications outside of astrophysics. Physics-inspired and physics-constrained neural networks [120] have shown promise in reducing the dimensionality of the underlying latent space of a network with an associated reduction in the size of data sets needed to train the network.…”

Developing a Victorious Strategy to the Second Strong Gravitational Lensing Data Challenge

“…Most of those techniques are based on image processing and/or Neural networks. In particular, several works have established that both traditional neural networks (Bom et al 2017;Estrada et al 2007) and deep neural networks (Petrillo et al 2019b;Jacobs et al 2019;Petrillo et al 2019a;Metcalf et al 2019;Lanusse et al 2018;Glazebrook et al 2017;Morgan et al 2021) can be used to identify lenses from non-lenses, with minimal human intervention. Although there is a certain intuition on how those techniques work, except when the methods explicitly take morphological features or colours rather than the raw image, it is not completely clear which features are used by Deep Learning algorithms in strong lensing identification.…”