arXiv subjects

Joshua S. Bloom

Publications and source records attributed to Joshua S. Bloom.

At least 55 records · Page 3Linked to original sources

arXiv2022

A very luminous jet from the disruption of a star by a massive black hole

Tidal disruption events (TDEs) are bursts of electromagnetic energy released when supermassive black holes (SMBHs) at the centers of galaxies violently disrupt a star that passes too close. TDEs provide a new window to study accretion onto SMBHs; in some rare cases, this accretion leads to launching of a relativistic jet, but the necessary conditions are not fully understood. The best studied jetted TDE to date is Swift J1644+57, which was discovered in gamma-rays, but was too obscured by dust to be seen at optical wavelengths. Here we report the optical discovery of AT2022cmc, a rapidly fading source at cosmological distance (redshift z=1.19325) whose unique lightcurve transitioned into a luminous plateau within days. Observations of a bright counterpart at other wavelengths, including X-rays, sub-millimeter, and radio, supports the interpretation of AT2022cmc as a jetted TDE containing a synchrotron "afterglow", likely launched by a SMBH with spin $a \gtrsim 0.3$. Using 4 years of Zwicky Transient Facility (ZTF) survey data, we calculate a rate of $0.02 ^{+ 0.04 }_{- 0.01 }$ Gpc$^{-3}$ yr$^{-1}$ for on-axis jetted TDEs based on the luminous, fast-fading red component, thus providing a measurement complementary to the rates derived from X-ray and radio observations. Correcting for the beaming angle effects, this rate confirms that about 1% of TDEs have relativistic jets. Optical surveys can use AT2022cmc as a prototype to unveil a population of jetted TDEs.

astro-ph.HE↗

arXiv2022

The Tidal Disruption Event AT2021ehb: Evidence of Relativistic Disk Reflection, and Rapid Evolution of the Disk-Corona System

We present X-ray, UV, optical, and radio observations of the nearby ($\approx78$ Mpc) tidal disruption event (TDE) AT2021ehb/ZTF21aanxhjv during its first 430 days of evolution. AT2021ehb occurs in the nucleus of a galaxy hosting a $\approx 10^{7}\,M_\odot$ black hole ($M_{\rm BH}$ inferred from host galaxy scaling relations). High-cadence Swift and NICER monitoring reveals a delayed X-ray brightening. The spectrum first undergoes a gradual ${\rm soft }\rightarrow{\rm hard}$ transition and then suddenly turns soft again within 3 days at $δt\approx 272$ days during which the X-ray flux drops by a factor of ten. In the joint NICER+NuSTAR observation ($δt =264$ days, harder state), we observe a prominent non-thermal component up to 30 keV and an extremely broad emission line in the iron K band. The bolometric luminosity of AT2021ehb reaches a maximum of $6.0^{+10.4}_{-3.8}\% L_{\rm Edd}$ when the X-ray spectrum is the hardest. During the dramatic X-ray evolution, no radio emission is detected, the UV/optical luminosity stays relatively constant, and the optical spectra are featureless. We propose the following interpretations: (i) the ${\rm soft }\rightarrow{\rm hard}$ transition may be caused by the gradual formation of a magnetically dominated corona; (ii) hard X-ray photons escape from the system along solid angles with low scattering optical depth ($\sim\,$a few) whereas the UV/optical emission is likely generated by reprocessing materials with much larger column density -- the system is highly aspherical; (iii) the abrupt X-ray flux drop may be triggered by the thermal-viscous instability in the inner accretion flow leading to a much thinner disk.

astro-ph.HE↗

arXiv2022

Cosmological Fast Optical Transients with the Zwicky Transient Facility: A Search for Dirty Fireballs

Dirty fireballs are a hypothesized class of relativistic massive-star explosions with an initial Lorentz factor $Γ_\mathrm{init}$ below the $Γ_\mathrm{init}\sim100$ required to produce a long-duration gamma-ray burst (LGRB), but which could still produce optical emission resembling LGRB afterglows. Here we present the results of a search for on-axis optical afterglows using the Zwicky Transient Facility (ZTF). Our search yielded seven optical transients that resemble on-axis LGRB afterglows in terms of their red colors ($g-r>0$ mag), faint host galaxy ($r>23$ mag), and rapid fading ($dr/dt>1$ mag/day). Spectroscopy of the transient emission within a few days of discovery established cosmological distances ($z=0.876$ to $z=2.9$) for six events, tripling the number of afterglows with redshift measurements discovered by optical surveys without a $γ$-ray trigger. Upon a retrospective search, four events (ZTF20abbiixp/AT2020kym, ZTF21aagwbjr/AT2021buv, ZTF21aakruew/AT2021cwd, ZTF21abfmpwn/AT2021qbd) turned out to have a likely associated LGRB (GRB200524A, GRB210204A, GRB210212B, GRB210610B), while three did not (ZTF20aajnksq/AT2020blt, ZTF21aaeyldq/AT2021any, ZTF21aayokph/AT2021lfa). Our search revealed no definitive new class of events: the simplest explanation for the apparently "orphan" events is that they were regular LGRBs missed by high-energy satellites due to detector sensitivity and duty cycle, although it is possible that they were intrinsically faint in $γ$-rays or viewed slightly off-axis. We rule out a scenario in which dirty fireballs have a similar energy per solid angle to LGRBs and are an order of magnitude more common. In addition, we set the first direct constraint on the ratio of the opening angles of the material producing $γ$-rays and the material producing early optical afterglow emission, finding that they must be comparable.

astro-ph.HE↗

arXiv2022

A Ubiquitous Unifying Degeneracy in Two-Body Microlensing Systems

While gravitational microlensing by planetary systems provides unique vistas on the properties of exoplanets, observations of a given 2-body microlensing event can often be interpreted with multiple distinct physical configurations. Such ambiguities are typically attributed to the close-wide and inner-outer types of degeneracies that arise from transformation invariances and symmetries of microlensing caustics. However, there remain unexplained inconsistencies between aforementioned theories and observations. Here, leveraging a fast machine learning inference framework, we present the discovery of the offset degeneracy, which concerns a magnification-matching behaviour on the lens-axis and is formulated independent of caustics. This offset degeneracy unifies the close-wide and inner-outer degeneracies, generalises to resonant topologies, and upon reanalysis, not only appears ubiquitous in previously published planetary events with 2-fold degenerate solutions, but also resolves prior inconsistencies. Our analysis demonstrates that degenerate caustics do not strictly result in degenerate magnifications and that the commonly invoked close-wide degeneracy essentially never arises in actual events. Moreover, it is shown that parameters in offset degenerate configurations are related by a simple expression. This suggests the existence of a deeper symmetry in the equations governing 2-body lenses than previously recognised.

astro-ph.EP↗

arXiv2022

HEALPix Alchemy: Fast All-Sky Geometry and Image Arithmetic in a Relational Database for Multimessenger Astronomy Brokers

Efficient searches for electromagnetic counterparts to gravitational wave, high-energy neutrino, and gamma-ray burst events demand rapid processing of image arithmetic and geometry set operations in a database to cross-match galaxy catalogs, observation footprints, and all-sky images. Here we introduce HEALPix Alchemy, an open-source, pure Python implementation of a set of methods that enables rapid all-sky geometry calculations. HEALPix Alchemy is built upon HEALPix, a spatial indexing strategy that is widely used in astronomical databases as well as the native format of LIGO-Virgo-KAGRA gravitational-wave sky localization maps. Our approach leverages new multirange types built into the PostgreSQL 14 database engine. This enables fast all-sky queries against probabilistic multimessenger event localizations and telescope survey footprints. Questions such as "What are the galaxies contained within the 90% credible region of an event?" and "What is the rank-ordered list of the fields within an observing footprint with the highest probability of containing the event?" can be performed in less than a few seconds on commodity hardware using off-the-shelf cloud-managed database implementations without server-side database extensions. Common queries scale roughly linearly with the number of telescope pointings. As the number of fields grows into the hundreds or thousands, HEALPix Alchemy is orders of magnitude faster than other implementations. HEALPix Alchemy is now used as the spatial geometry engine within SkyPortal, which forms the basis of the Zwicky Transient Facility transient marshal, called Fritz.

astro-ph.IM↗

arXiv2022

Target of Opportunity Observations of Gravitational Wave Events with Vera C. Rubin Observatory

The discovery of the electromagnetic counterpart to the binary neutron star merger GW170817 has opened the era of gravitational-wave multi-messenger astronomy. Rapid identification of the optical/infrared kilonova enabled a precise localization of the source, which paved the way to deep multi-wavelength follow-up and its myriad of related science results. Fully exploiting this new territory of exploration requires the acquisition of electromagnetic data from samples of neutron star mergers and other gravitational wave sources. After GW170817, the frontier is now to map the diversity of kilonova properties and provide more stringent constraints on the Hubble constant, and enable new tests of fundamental physics. The Vera C. Rubin Observatory's Legacy Survey of Space and Time (LSST) can play a key role in this field in the 2020s, when an improved network of gravitational-wave detectors is expected to reach a sensitivity that will enable the discovery of a high rate of merger events involving neutron stars (about tens per year) out to distances of several hundred Mpc. We design comprehensive target-of-opportunity observing strategies for follow-up of gravitational-wave triggers that will make the Rubin Observatory the premier instrument for discovery and early characterization of neutron star and other compact object mergers, and yet unknown classes of gravitational wave events.

astro-ph.HE↗

arXiv2022

In search of short gamma-ray burst optical counterpart with the Zwicky Transient Facility

The Fermi Gamma-ray Burst Monitor (GBM) triggers on-board in response to $\sim$ 40 short gamma-ray bursts (SGRBs) per year; however, their large localization regions have made the search for optical counterparts a challenging endeavour. We have developed and executed an extensive program with the wide field of view of the Zwicky Transient Facility (ZTF) camera, mounted on the Palomar 48 inch Oschin telescope (P48), to perform target-of-opportunity (ToO) observations on 10 Fermi-GBM SGRBs during 2018 and 2020-2021. Bridging the large sky areas with small field of view optical telescopes in order to track the evolution of potential candidates, we look for the elusive SGRB afterglows and kilonovae (KNe) associated with these high-energy events. No counterpart has yet been found, even though more than 10 ground based telescopes, part of the Global Relay of Observatories Watching Transients Happen (GROWTH) network, have taken part in these efforts. The candidate selection procedure and the follow-up strategy have shown that ZTF is an efficient instrument for searching for poorly localized SGRBs, retrieving a reasonable number of candidates to follow-up and showing promising capabilities as the community approaches the multi-messenger era. Based on the median limiting magnitude of ZTF, our searches would have been able to retrieve a GW170817-like event up to $\sim$ 200 Mpc and SGRB afterglows to z = 0.16 or 0.4, depending on the assumed underlying energy model. Future ToOs will expand the horizon to z = 0.2 and 0.7 respectively.

astro-ph.HE↗

arXiv2021

The Gravity Collective: A Search for the Electromagnetic Counterpart to the Neutron Star-Black Hole Merger GW190814

We present optical follow-up imaging obtained with the Katzman Automatic Imaging Telescope, Las Cumbres Observatory Global Telescope Network, Nickel Telescope, Swope Telescope, and Thacher Telescope of the LIGO/Virgo gravitational wave (GW) signal from the neutron star-black hole (NSBH) merger GW190814. We searched the GW190814 localization region (19 deg$^{2}$ for the 90th percentile best localization), covering a total of 51 deg$^{2}$ and 94.6% of the two-dimensional localization region. Analyzing the properties of 189 transients that we consider as candidate counterparts to the NSBH merger, including their localizations, discovery times from merger, optical spectra, likely host-galaxy redshifts, and photometric evolution, we conclude that none of these objects are likely to be associated with GW190814. Based on this finding, we consider the likely optical properties of an electromagnetic counterpart to GW190814, including possible kilonovae and short gamma-ray burst afterglows. Using the joint limits from our follow-up imaging, we conclude that a counterpart with an $r$-band decline rate of 0.68 mag day$^{-1}$, similar to the kilonova AT 2017gfo, could peak at an absolute magnitude of at most $-17.8$ mag (50% confidence). Our data are not constraining for ''red'' kilonovae and rule out ''blue'' kilonovae with $M>0.5 M_{\odot}$ (30% confidence). We strongly rule out all known types of short gamma-ray burst afterglows with viewing angles $<$17$^{\circ}$ assuming an initial jet opening angle of $\sim$$5.2^{\circ}$ and explosion energies and circumburst densities similar to afterglows explored in the literature. Finally, we explore the possibility that GW190814 merged in the disk of an active galactic nucleus, of which we find four in the localization region, but we do not find any candidate counterparts among these sources.

astro-ph.HE↗

arXiv2021

Classification of Periodic Variable Stars with Novel Cyclic-Permutation Invariant Neural Networks

Neural networks (NNs) have been shown to be competitive against state-of-the-art feature engineering and random forest (RF) classification of periodic variable stars. Although previous work utilising NNs has made use of periodicity by period folding multiple-cycle time-series into a single cycle -- from time-space to phase-space -- no approach to date has taken advantage of the fact that network predictions should be invariant to the initial phase of the period-folded sequence. Initial phase is exogenous to the physical origin of the variability and should thus be factored out. Here, we present cyclic-permutation invariant networks, a novel class of NNs for which invariance to phase shifts is guaranteed through polar coordinate convolutions, which we implement by means of "Symmetry Padding." Across three different datasets of variable star light curves, we show that two implementations of the cyclic-permutation invariant network: the iTCN and the iResNet, consistently outperform non-invariant baselines and reduce overall error rates by between 4% to 22%. Over a 10-class OGLE-III sample, the iTCN/iResNet achieves an average per-class accuracy of 93.4%/93.3%, compared to RNN/RF accuracies of 70.5%/89.5% in a recent study using the same data. Finding improvement on a non-astronomy benchmark, we suggest that the methodology introduced here should also be applicable to a wide range of science domains where periodic data abounds due to physical symmetries.

astro-ph.IM↗

arXiv2021

Real-Time Likelihood-Free Inference of Roman Binary Microlensing Events with Amortized Neural Posterior Estimation

Fast and automated inference of binary-lens, single-source (2L1S) microlensing events with sampling-based Bayesian algorithms (e.g., Markov Chain Monte Carlo; MCMC) is challenged on two fronts: high computational cost of likelihood evaluations with microlensing simulation codes, and a pathological parameter space where the negative-log-likelihood surface can contain a multitude of local minima that are narrow and deep. Analysis of 2L1S events usually involves grid searches over some parameters to locate approximate solutions as a prerequisite to posterior sampling, an expensive process that often requires human-in-the-loop domain expertise. As the next-generation, space-based microlensing survey with the Roman Space Telescope is expected to yield thousands of binary microlensing events, a new fast and automated method is desirable. Here, we present a likelihood-free inference (LFI) approach named amortized neural posterior estimation, where a neural density estimator (NDE) learns a surrogate posterior $\hat{p}(θ|x)$ as an observation-parametrized conditional probability distribution, from pre-computed simulations over the full prior space. Trained on 291,012 simulated Roman-like 2L1S simulations, the NDE produces accurate and precise posteriors within seconds for any observation within the prior support without requiring a domain expert in the loop, thus allowing for real-time and automated inference. We show that the NDE also captures expected posterior degeneracies. The NDE posterior could then be refined into the exact posterior with a downstream MCMC sampler with minimal burn-in steps.

astro-ph.IM↗

arXiv2021

Automating Inference of Binary Microlensing Events with Neural Density Estimation

Automated inference of binary microlensing events with traditional sampling-based algorithms such as MCMC has been hampered by the slowness of the physical forward model and the pathological likelihood surface. Current analysis of such events requires both expert knowledge and large-scale grid searches to locate the approximate solution as a prerequisite to MCMC posterior sampling. As the next generation, space-based microlensing survey with the Roman Space Observatory is expected to yield thousands of binary microlensing events, a new scalable and automated approach is desired. Here, we present an automated inference method based on neural density estimation (NDE). We show that the NDE trained on simulated Roman data not only produces fast, accurate, and precise posteriors but also captures expected posterior degeneracies. A hybrid NDE-MCMC framework can further be applied to produce the exact posterior.

astro-ph.IM↗

arXiv2020

Modelling the prompt optical emission of GRB 180325A: the evolution of a spike from the optical to gamma-rays

The transition from prompt to the afterglow emission is one of the most exciting and least understood phases in gamma-ray bursts (GRBs). Correlations among optical, X-ray and gamma-ray emission in GRBs have been explored, to attempt to answer whether the earliest optical emission comes from internal and/or external shocks. We present optical photometric observations of GRB 180325A collected with the TAROT and RATIR ground-based telescopes. These observations show two strong optical flashes with separate peaks at $\sim50\;$s and $\sim120\;$s, followed by a temporally extended optical emission. We also present X-rays and gamma-ray observations of GRB 180325A, detected by the Burst Alert Telescope (BAT) and X-ray Telescope (XRT), on the Neil Gehrels Swift observatory, which both observed a narrow flash at $\sim80\;$s. We show that the prompt gamma- and X-ray early emission shares similar temporal and spectral features consistent with internal dissipation within the relativistic outflow (e.g. by internal shocks or magnetic reconnection), while the early optical flashes are likely generated by the reverse shock that decelerates the ejecta as it sweeps up the external medium.

astro-ph.HE↗

arXiv2020

Informing the Cataclysmic Variable Donor Sequence from Gaia DR2 Color-Magnitude and Inferred Variability Metrics

Short-period cataclysmic variables (spCVs), with orbital periods below the period gap ($P_{orb}$ < 2 hr), offer insight into the evolutionary models of CVs and can serve as strong emitters of gravitational waves (GWs). To identify new spCV candidates, we crossmatch a catalog of known CVs to sources with robust parallaxes in the Gaia second data release (DR2). We uncover and fit an apparently monotonic relationship between the color--absolute-magnitude diagram (CMD) position and $P_{orb}$ of these CVs, revealed in DR2. To supplement this relation, we develop a method for identifying sources with large photometric variability, a characteristic trait of spCVs. Using all available Gaia light curves, we construct a machine-learned regression model to predict variability metrics for sources in the CMD locus of known spCVs based solely on time-averaged covariates present in DR2. Using this approach we identify 3,253 candidate spCVs, of which $\sim$95% are previously unknown. Inspection of archival SDSS spectra of these candidates suggests that $>$82% are likely to be spCVs: a noticeably higher recovery rate than previous light curve searches, which bias toward active systems. We obtain optical spectra of 9 new systems at Lick Observatory and confirm that all objects are CV systems. We measure $P_{orb}$ for 7 systems using archival Gaia and Palomar Transient Factory light curves, 3 of which do not have previous $P_{orb}$ measurements. We use the CMD-$P_{orb}$ relation to infer the detectability of these systems to the upcoming LISA mission, and find that six sources may be coherent LISA verification binaries, with an estimated SNR > 5 in the 4 yr mission. This paper demonstrates that the time-averaged Gaia catalog is a powerful tool in the methodical discovery and characterization of time-varying objects, making it complementary to missions like ZTF, TESS, and the Vera Rubin LSST.

astro-ph.SR↗

arXiv2020

The Palomar Transient Factory Core-Collapse Supernova Host-Galaxy Sample. I. Host-Galaxy Distribution Functions and Environment-Dependence of CCSNe

Several thousand core-collapse supernovae (CCSNe) of different flavors have been discovered so far. However, identifying their progenitors has remained an outstanding open question in astrophysics. Studies of SN host galaxies have proven to be powerful in providing constraints on the progenitor populations. In this paper, we present all CCSNe detected between 2009 and 2017 by the Palomar Transient Factory. This sample includes 888 SNe of 12 distinct classes out to redshift $z\approx1$. We present the photometric properties of their host galaxies from the far-ultraviolet to the mid-infrared and model the host-galaxy spectral energy distributions to derive physical properties. The galaxy mass functions of Type Ic, Ib, IIb, II, and IIn SNe ranges from $10^{5}$ to $10^{11.5}~M_\odot$, probing the entire mass range of star-forming galaxies down to the least-massive star-forming galaxies known. Moreover, the galaxy mass distributions are consistent with models of star-formation-weighted mass functions. Regular CCSNe are hence direct tracers of star formation. Small but notable differences exist between some of the SN classes. Type Ib/c SNe prefer galaxies with slightly higher masses (i.e., higher metallicities) and star-formation rates than Type IIb and II SNe. These differences are less pronounced than previously thought. H-poor SLSNe and SNe~Ic-BL are scarce in galaxies above $10^{10}~M_\odot$. Their progenitors require environments with metallicities of $<0.4$ and $<1$ solar, respectively. In addition, the hosts of H-poor SLSNe are dominated by a younger stellar population than all other classes of CCSNe. Our findings corroborate the notion that low-metallicity \textit{and} young age play an important role in the formation of SLSN progenitors.

astro-ph.GA↗

arXiv2020

Kilonova Luminosity Function Constraints based on Zwicky Transient Facility Searches for 13 Neutron Star Mergers

We present a systematic search for optical counterparts to 13 gravitational wave (GW) triggers involving at least one neutron star during LIGO/Virgo's third observing run. We searched binary neutron star (BNS) and neutron star black hole (NSBH) merger localizations with the Zwicky Transient Facility (ZTF) and undertook follow-up with the Global Relay of Observatories Watching Transients Happen (GROWTH) collaboration. The GW triggers had a median localization of 4480 deg^2, median distance of 267 Mpc and false alarm rates ranging from 1.5 to 1e-25 per yr. The ZTF coverage had a median enclosed probability of 39%, median depth of 20.8mag, and median response time of 1.5 hr. The O3 follow-up by the GROWTH team comprised 340 UVOIR photometric points, 64 OIR spectra, and 3 radio. We find no promising kilonova (radioactivity-powered counterpart) and we convert the upper limits to constrain the underlying kilonova luminosity function. Assuming that all kilonovae are at least as luminous as GW170817 at discovery (-16.1mag), we calculate our joint probability of detecting zero kilonovae is only 4.2%. If we assume that all kilonovae are brighter than -16.6mag (extrapolated peak magnitude of GW170817) and fade at 1 mag/day (similar to GW170817), the joint probability of zero detections is 7%. If we separate the NSBH and BNS populations, the joint probability of zero detections, assuming all kilonovae are brighter than -16.6mag, is 9.7% for NSBH and 7.9% for BNS mergers. Moreover, <57% (<89%) of putative kilonovae could be brighter than -16.6mag assuming flat (fading) evolution, at 90% confidence. If we further account for the online terrestrial probability for each GW trigger, we find that <68% of putative kilonovae could be brighter than -16.6mag. Comparing to model grids, we find that some kilonovae must have Mej < 0.03 Msun or Xlan>1e-4 or phi>30deg to be consistent with our limits. (Abridged)

astro-ph.HE↗

arXiv2020

deepSIP: Linking Type Ia Supernova Spectra to Photometric Quantities with Deep Learning

We present {\tt deepSIP} (deep learning of Supernova Ia Parameters), a software package for measuring the phase and -- for the first time using deep learning -- the light-curve shape of a Type Ia supernova (SN~Ia) from an optical spectrum. At its core, {\tt deepSIP} consists of three convolutional neural networks trained on a substantial fraction of all publicly-available low-redshift SN~Ia optical spectra, onto which we have carefully coupled photometrically-derived quantities. We describe the accumulation of our spectroscopic and photometric datasets, the cuts taken to ensure quality, and our standardised technique for fitting light curves. These considerations yield a compilation of 2754 spectra with photometrically characterised phases and light-curve shapes. Though such a sample is significant in the SN community, it is small by deep-learning standards where networks routinely have millions or even billions of free parameters. We therefore introduce a data-augmentation strategy that meaningfully increases the size of the subset we allocate for training while prioritising model robustness and telescope agnosticism. We demonstrate the effectiveness of our models by deploying them on a sample unseen during training and hyperparameter selection, finding that Model~I identifies spectra that have a phase between $-10$ and 18\,d and light-curve shape, parameterised by $Δm_{15}$, between 0.85 and 1.55\,mag with an accuracy of 94.6\%. For those spectra that do fall within the aforementioned region in phase--$Δm_{15}$ space, Model~II predicts phases with a root-mean-square error (RMSE) of 1.00\,d and Model~III predicts $Δm_{15}$ values with an RMSE of 0.068\,mag.

astro-ph.IM↗

arXiv2020

On Neural Architectures for Astronomical Time-series Classification with Application to Variable Stars

Despite the utility of neural networks (NNs) for astronomical time-series classification, the proliferation of learning architectures applied to diverse datasets has thus far hampered a direct intercomparison of different approaches. Here we perform the first comprehensive study of variants of NN-based learning and inference for astronomical time-series, aiming to provide the community with an overview on relative performance and, hopefully, a set of best-in-class choices for practical implementations. In both supervised and self-supervised contexts, we study the effects of different time-series-compatible layer choices, namely the dilated temporal convolutional neural network (dTCNs), Long-Short Term Memory (LSTM) NNs, Gated Recurrent Units (GRUs) and temporal convolutional NNs (tCNNs). We also study the efficacy and performance of encoder-decoder (i.e., autoencoder) networks compared to direct classification networks, different pathways to include auxiliary (non-time-series) metadata, and different approaches to incorporate multi-passband data (i.e., multiple time-series per source). Performance---applied to a sample of 17,604 variable stars from the MACHO survey across 10 imbalanced classes---is measured in training convergence time, classification accuracy, reconstruction error, and generated latent variables. We find that networks with Recurrent NN (RNNs) generally outperform dTCNs and, in many scenarios, yield to similar accuracy as tCNNs. In learning time and memory requirements, convolution-based layers are more performant. We conclude by discussing the advantages and limitations of deep architectures for variable star classification, with a particular eye towards next-generation surveys such as LSST, WFIRST and ZTF2.

astro-ph.IM↗

arXiv2020

Deep Generative Modeling of Periodic Variable Stars Using Physical Parameters

The ability to generate physically plausible ensembles of variable sources is critical to the optimization of time-domain survey cadences and the training of classification models on datasets with few to no labels. Traditional data augmentation techniques expand training sets by reenvisioning observed exemplars, seeking to simulate observations of specific training sources under different (exogenous) conditions. Unlike fully theory-driven models, these approaches do not typically allow principled interpolation nor extrapolation. Moreover, the principal drawback of theory-driven models lies in the prohibitive computational cost of simulating source observables from {\it ab initio} parameters. In this work, we propose a computationally tractable machine learning approach to generate realistic light curves of periodic variables capable of integrating physical parameters and variability classes as inputs. Our deep generative model, inspired by the Transparent Latent Space Generative Adversarial Networks (TL-GANs), uses a Variational Autoencoder (VAE) architecture with Temporal Convolutional Network (TCN) layers, trained using the \hbox{OGLE-III} optical light curves and physical characteristics (e.g., effective temperature and absolute magnitude) from Gaia DR2. A test using the temperature-shape relationship of RR\,Lyrae demonstrates the efficacy of our generative "Physics-Enhanced Latent Space VAE" (PELS-VAE) model. Such deep generative models, serving as non-linear non-parametric emulators, present a novel tool for astronomers to create synthetic time series over arbitrary cadences.

astro-ph.IM↗