-
YOLO-CIANNA: Galaxy detection with deep learning in radio data. I. A new YOLO-inspired source detection method applied to the SKAO SDC1
Authors:
D. Cornu,
P. Salomé,
B. Semelin,
A. Marchal,
J. Freundlich,
S. Aicardi,
X. Lu,
G. Sainton,
F. Mertens,
F. Combes,
C. Tasse
Abstract:
The upcoming Square Kilometer Array (SKA) will set a new standard regarding data volume generated by an astronomical instrument, which is likely to challenge widely adopted data analysis tools that scale inadequately with the data size. This study aims to develop a new source detection and characterization method for massive radio astronomical datasets by adapting modern deep-learning object detec…
▽ More
The upcoming Square Kilometer Array (SKA) will set a new standard regarding data volume generated by an astronomical instrument, which is likely to challenge widely adopted data analysis tools that scale inadequately with the data size. This study aims to develop a new source detection and characterization method for massive radio astronomical datasets by adapting modern deep-learning object detection techniques. These approaches have proved their efficiency on complex computer vision tasks, and we seek to identify their specific strengths and weaknesses when applied to astronomical data. We introduce YOLO-CIANNA, a highly customized deep-learning object detector designed specifically for astronomical datasets. This paper presents the method and describes all the low-level adaptations required to address the specific challenges of radio-astronomical images. We demonstrate this method's capabilities using simulated 2D continuum images from the SKAO SDC1 dataset. Our method outperforms every other published result on the specific SDC1 dataset. Using the SDC1 metric, we improve the challenge-winning score by +139\% and the score of the only other post-challenge participation by +61\%. Our catalog has a detection purity of 94\% while detecting 40 to 60 \% more sources than previous top-score results. The trained model can also be forced to reach 99\% purity in post-process and still detect 10 to 30\% more sources than the other top-score methods. It is also capable of real-time detection, with a peak prediction speed of 500 images of 512x512 pixels per second on a single GPU. YOLO-CIANNA achieves state-of-the-art detection and characterization results on the simulated SDC1 dataset. The method is open source and included in the wider CIANNA framework. We provide scripts to train and apply this method to the SDC1 dataset in the CIANNA repository.
△ Less
Submitted 8 February, 2024;
originally announced February 2024.
-
Martian time-series unraveled: A multi-scale nested approach with factorial variational autoencoders
Authors:
Ali Siahkoohi,
Rudy Morel,
Randall Balestriero,
Erwan Allys,
Grégory Sainton,
Taichi Kawamura,
Maarten V. de Hoop
Abstract:
Unsupervised source separation involves unraveling an unknown set of source signals recorded through a mixing operator, with limited prior knowledge about the sources, and only access to a dataset of signal mixtures. This problem is inherently ill-posed and is further challenged by the variety of timescales exhibited by sources. Existing methods typically rely on a preselected window size that det…
▽ More
Unsupervised source separation involves unraveling an unknown set of source signals recorded through a mixing operator, with limited prior knowledge about the sources, and only access to a dataset of signal mixtures. This problem is inherently ill-posed and is further challenged by the variety of timescales exhibited by sources. Existing methods typically rely on a preselected window size that determines their operating timescale, limiting their capacity to handle multi-scale sources. To address this issue, we propose an unsupervised multi-scale clustering and source separation framework by leveraging wavelet scattering spectra that provide a low-dimensional representation of stochastic processes, capable of distinguishing between different non-Gaussian stochastic processes. Nested within this representation space, we develop a factorial Gaussian-mixture variational autoencoder that is trained to (1) probabilistically cluster sources at different timescales and (2) independently sample scattering spectra representations associated with each cluster. As the final stage, using samples from each cluster as prior information, we formulate source separation as an optimization problem in the wavelet scattering spectra representation space, aiming to separate sources in the time domain. When applied to the entire seismic dataset recorded during the NASA InSight mission on Mars, containing sources varying greatly in timescale, our multi-scale nested approach proves to be a powerful tool for disentangling such different sources, e.g., minute-long transient one-sided pulses (known as ``glitches'') and structured ambient noises resulting from atmospheric activities that typically last for tens of minutes. These results provide an opportunity to conduct further investigations into the isolated sources related to atmospheric-surface interactions, thermal relaxations, and other complex phenomena.
△ Less
Submitted 19 February, 2024; v1 submitted 25 May, 2023;
originally announced May 2023.
-
Unearthing InSights into Mars: Unsupervised Source Separation with Limited Data
Authors:
Ali Siahkoohi,
Rudy Morel,
Maarten V. de Hoop,
Erwan Allys,
Grégory Sainton,
Taichi Kawamura
Abstract:
Source separation involves the ill-posed problem of retrieving a set of source signals that have been observed through a mixing operator. Solving this problem requires prior knowledge, which is commonly incorporated by imposing regularity conditions on the source signals, or implicitly learned through supervised or unsupervised methods from existing data. While data-driven methods have shown great…
▽ More
Source separation involves the ill-posed problem of retrieving a set of source signals that have been observed through a mixing operator. Solving this problem requires prior knowledge, which is commonly incorporated by imposing regularity conditions on the source signals, or implicitly learned through supervised or unsupervised methods from existing data. While data-driven methods have shown great promise in source separation, they often require large amounts of data, which rarely exists in planetary space missions. To address this challenge, we propose an unsupervised source separation scheme for domains with limited data access that involves solving an optimization problem in the wavelet scattering covariance representation space$\unicode{x2014}$an interpretable, low-dimensional representation of stationary processes. We present a real-data example in which we remove transient, thermally-induced microtilts$\unicode{x2014}$known as glitches$\unicode{x2014}$from data recorded by a seismometer during NASA's InSight mission on Mars. Thanks to the wavelet scattering covariances' ability to capture non-Gaussian properties of stochastic processes, we are able to separate glitches using only a few glitch-free data snippets.
△ Less
Submitted 31 May, 2023; v1 submitted 27 January, 2023;
originally announced January 2023.
-
Spectroscopic observations of eight supernovae at intermediate redshift
Authors:
C. Balland,
M. Mouchet,
R. Amanullah,
P. Astier,
S. Fabbro,
G. Folatelli,
G. Garavini,
A. Goobar,
D. Hardin,
M. J. Irwin,
R. G. McMahon,
A. -M. Mourao,
S. Nobili,
R. Pain,
R. Pascoal,
J. Raux,
G. Sainton,
K. Schahmaneche,
N. A. Walton
Abstract:
We present spectra of six Type Ia and two Type II supernovae obtained in June 2002 at the William Herschel Telescope during a search for Type Ia supernovae (SNIa) at intermediate redshift. Supernova type identification and phase determination are performed using a fitting technique based on a Xi2 minimization against a series of model templates. The spectra range from z=0.033 to z=0.328, includi…
▽ More
We present spectra of six Type Ia and two Type II supernovae obtained in June 2002 at the William Herschel Telescope during a search for Type Ia supernovae (SNIa) at intermediate redshift. Supernova type identification and phase determination are performed using a fitting technique based on a Xi2 minimization against a series of model templates. The spectra range from z=0.033 to z=0.328, including one spectroscopically underluminous SNIa at z=0.033. This set of spectra significantly increases the sample of well-observed type SNIa supernovae available in the range 0.15< z <0.35. Together with the twelve supernovae observed by our team in 1999 in the same redshift range, they form an homogeneous sample of seventeen type Ia supernovae with comparable signal-to-noise ratio and regular phase sampling in a still largely unexplored region of the redshift space.
△ Less
Submitted 26 March, 2007;
originally announced March 2007.
-
Quantitative comparison between Type Ia supernova spectra at low and high redshifts: A case study
Authors:
G. Garavini,
G. Folatelli,
S. Nobili,
G. Aldering,
R. Amanullah,
P. Antilogus,
P. Astier,
G. Blanc,
J. Bronder,
M. S. Burns,
A. Conley,
S. E. Deustua,
M. Doi,
S. Fabbro,
V. Fadeyev,
R. Gibbons,
G. Goldhaber,
A. Goobar,
D. E. Groom,
I. Hook,
D. A. Howell,
N. Kashikawa,
A. G. Kim,
M. Kowalski,
N. Kuznetsova
, et al. (22 additional authors not shown)
Abstract:
We develop a method to measure the strength of the absorption features in Type Ia supernova (SN Ia) spectra and use it to make a quantitative comparison between the spectra of Type Ia supernovae at low and high redshifts. In this case study, we apply the method to 12 high-redshift (0.212 < z < 0.912) SNe Ia observed by the Supernova Cosmology Project . Through measurements of the strengths of th…
▽ More
We develop a method to measure the strength of the absorption features in Type Ia supernova (SN Ia) spectra and use it to make a quantitative comparison between the spectra of Type Ia supernovae at low and high redshifts. In this case study, we apply the method to 12 high-redshift (0.212 < z < 0.912) SNe Ia observed by the Supernova Cosmology Project . Through measurements of the strengths of these features and of the blueshift of the absorption minimum in Ca II H&K, we show that the spectra of the high-redshift SNe Ia are quantitatively similar to spectra of nearby SNe Ia (z < 0.15). One supernova in our high redshift sample, SN 2002fd at z=0.279, is found to have spectral characteristics that are associated with peculiar SN 1991T/SN 1999aa-like supernovae.
△ Less
Submitted 24 March, 2007;
originally announced March 2007.
-
Spectra of High-Redshift Type Ia Supernovae and a Comparison with their Low-Redshift Counterparts
Authors:
I. M. Hook,
D. A. Howell,
G. Aldering,
R. Amanullah,
M. S. Burns,
A. Conley,
S. E. Deustua,
R. Ellis,
S. Fabbro,
V. Fadeyev,
G. Folatelli,
G. Garavini,
R. Gibbons,
G. Goldhaber,
A. Goobar,
D. E. Groom,
A. G. Kim,
R. A. Knop,
M. Kowalski,
C. Lidman,
S. Nobili,
P. E. Nugent,
R. Pain,
C. R. Pennypacker,
S. Perlmutter
, et al. (10 additional authors not shown)
Abstract:
We present spectra for 14 high-redshift (0.17 < z < 0.83) supernovae, which were discovered by the Supernova Cosmology Project as part of a campaign to measure cosmological parameters. The spectra are used to determine the redshift and classify the supernova type, essential information if the supernovae are to be used for cosmological studies. Redshifts were derived either from the spectrum of t…
▽ More
We present spectra for 14 high-redshift (0.17 < z < 0.83) supernovae, which were discovered by the Supernova Cosmology Project as part of a campaign to measure cosmological parameters. The spectra are used to determine the redshift and classify the supernova type, essential information if the supernovae are to be used for cosmological studies. Redshifts were derived either from the spectrum of the host galaxy or from the spectrum of the supernova itself. We present evidence that these supernovae are of Type Ia by matching to spectra of nearby supernovae. We find that the dates of the spectra relative to maximum light determined from this fitting process are consistent with the dates determined from the photometric light curves, and moreover the spectral time-sequence for SNe Type Ia at low and high redshift is indistinguishable. We also show that the expansion velocities measured from blueshifted CaHK are consistent with those measured for low-redshift Type Ia supernovae. From these first-level quantitative comparisons we find no evidence for evolution in SNIa properties between these low- and high-redshift samples. Thus even though our samples may not be complete, we conclude that there is a population of SNe Ia at high redshift whose spectral properties match those at low redshift.
△ Less
Submitted 2 September, 2005;
originally announced September 2005.
-
Spectroscopy of twelve Type Ia supernovae at intermediate redshift
Authors:
C. Balland,
M. Mouchet,
R. Pain,
N. A. Walton,
R. Amanullah,
P. Astier,
R. S. Ellis,
S. Fabbro,
A. Goobar,
D. Hardin,
I. M. Hook,
M. J. Irwin,
R. M. McMahon,
J. M. Mendez,
P. Ruiz-Lapuente,
G. Sainton,
K. Schahmaneche,
V. Stanishev
Abstract:
We present spectra of twelve Type Ia supernovae obtained in 1999 at the William Herschel Telescope and the Nordic Optical Telescope during a search for Type Ia supernovae (SN Ia) at intermediate redshift. The spectra range from z=0.178 to z=0.493, including five high signal-to-noise ratio SN Ia spectra in the still largely unexplored range 0.15 < z < 0.3. Most of the spectra were obtained before…
▽ More
We present spectra of twelve Type Ia supernovae obtained in 1999 at the William Herschel Telescope and the Nordic Optical Telescope during a search for Type Ia supernovae (SN Ia) at intermediate redshift. The spectra range from z=0.178 to z=0.493, including five high signal-to-noise ratio SN Ia spectra in the still largely unexplored range 0.15 < z < 0.3. Most of the spectra were obtained before or around restframe B-band maximum light. None of them shows the peculiar spectral features found in low-redshift over- or under-luminous SN Ia. Expansion velocities of characteristic spectral absorption features such as SiII at 6355 angs., SII at 5640 angs. and CaII at 3945 angs. are found consistent with their low-z SN Ia counterparts.
△ Less
Submitted 29 July, 2005;
originally announced July 2005.
-
Restframe I-band Hubble diagram for type Ia supernovae up to redshift z ~0.5
Authors:
The Supernova Cosmology Project,
:,
S. Nobili,
R. Amanullah,
G. Garavini,
A. Goobar,
C. Lidman,
V. Stanishev,
G. Aldering,
P. Antilogus,
P. Astier,
M. S. Burns,
A. Conley,
S. E. Deustua,
R. Ellis,
S. Fabbro,
V. Fadeyev,
G. Folatelli,
R. Gibbons,
G. Goldhaber,
D. E. Groom,
I. Hook,
D. A. Howell,
A. G. Kim,
R. A. Knop
, et al. (13 additional authors not shown)
Abstract:
We present a novel technique for fitting restframe I-band light curves on a data set of 42 Type Ia supernovae (SNe Ia). Using the result of the fit, we construct a Hubble diagram with 26 SNe from the subset at 0.01< z<0.1. Adding two SNe at z~0.5 yields results consistent with a flat Lambda-dominated``concordance universe'' ($Ω_M,Ω_Λ$)=(0.25,0.75). For one of these, SN 2000fr, new near infrared…
▽ More
We present a novel technique for fitting restframe I-band light curves on a data set of 42 Type Ia supernovae (SNe Ia). Using the result of the fit, we construct a Hubble diagram with 26 SNe from the subset at 0.01< z<0.1. Adding two SNe at z~0.5 yields results consistent with a flat Lambda-dominated``concordance universe'' ($Ω_M,Ω_Λ$)=(0.25,0.75). For one of these, SN 2000fr, new near infrared data are presented. The high redshift supernova NIR data are also used to test for systematic effects in the use of SNe Ia as distance estimators. A flat, Lambda=0, universe where the faintness of supernovae at z~0.5 is due to grey dust homogeneously distributed in the intergalactic medium is disfavoured based on the high-z Hubble diagram using this small data-set. However, the uncertainties are large and no firm conclusion may be drawn. We explore the possibility of setting limits on intergalactic dust based on B-I and B-V colour measurements, and conclude that about 20 well measured SNe are needed to give statistically significant results. We also show that the high redshift restframe I-band data points are better fit by light curve templates that show a prominent second peak, suggesting that they are not intrinsically underluminous.
△ Less
Submitted 6 April, 2005;
originally announced April 2005.
-
Spectroscopic confirmation of high-redshift supernovae with the ESO VLT
Authors:
C. Lidman,
D. A. Howell,
G. Folatelli,
G. Garavini,
S. Nobili,
G. Aldering,
R. Amanullah,
P. Antilogus,
P. Astier,
G. Blanc,
M. S. Burns,
A. Conley,
S. E. Deustua,
M. Doi,
R. Ellis,
S. Fabbro,
V. Fadeyev,
R. Gibbons,
G. Goldhaber,
A. Goobar,
D. E. Groom,
I. Hook,
N. Kashikawa,
A. G. Kim,
R. A. Knop
, et al. (22 additional authors not shown)
Abstract:
We present VLT FORS1 and FORS2 spectra of 39 candidate high-redshift supernovae that were discovered as part of a cosmological study using Type Ia supernovae (SNe Ia) over a wide range of redshifts. From the spectra alone, 20 candidates are spectrally classified as SNe Ia with redshifts ranging from z=0.212 to z=1.181. Of the remaining 19 candidates, 1 might be a Type II supernova and 11 exhibit…
▽ More
We present VLT FORS1 and FORS2 spectra of 39 candidate high-redshift supernovae that were discovered as part of a cosmological study using Type Ia supernovae (SNe Ia) over a wide range of redshifts. From the spectra alone, 20 candidates are spectrally classified as SNe Ia with redshifts ranging from z=0.212 to z=1.181. Of the remaining 19 candidates, 1 might be a Type II supernova and 11 exhibit broad supernova-like spectral features and/or have supernova-like light curves. The candidates were discovered in 8 separate ground-based searches. In those searches in which SNe Ia at z ~ 0.5 were targeted, over 80% of the observed candidates were spectrally classified as SNe Ia. In those searches in which SNe Ia with z > 1 were targeted, 4 candidates with z > 1 were spectrally classified as SNe Ia and later followed with ground and space based observatories. We present the spectra of all candidates, including those that could not be spectrally classified as supernova.
△ Less
Submitted 21 October, 2004;
originally announced October 2004.
-
Extracting clean supernova spectra
Authors:
S. Blondin,
J. R. Walsh,
B. Leibundgut,
G. Sainton
Abstract:
We use a new technique to extract the spectrum of a supernova from that of the contaminating background of its host galaxy, and apply it to the specific case of high-redshift Type Ia supernova (SN Ia) spectroscopy. The algorithm is based on a two-channel iterative technique employing the Richardson-Lucy restoration method and is implemented in the IRAF code 'specinholucy'. We run the code both o…
▽ More
We use a new technique to extract the spectrum of a supernova from that of the contaminating background of its host galaxy, and apply it to the specific case of high-redshift Type Ia supernova (SN Ia) spectroscopy. The algorithm is based on a two-channel iterative technique employing the Richardson-Lucy restoration method and is implemented in the IRAF code 'specinholucy'. We run the code both on simulated (SN Ia at z=0.5 embedded in a bright host galaxy) and observed (SNe Ia at various phases up to z=0.236) data taken with VLT+FORS1 and show the advantages of using such a deconvolution technique in comparison with less elaborate methods. This paper is motivated by the need for optimal supernova spectroscopic data reduction in order to make meaningful comparisons between the low and high-redshift SN Ia samples. This may reveal subtle evolutionary and systematic effects that could depend on redshift and bias the cosmological results derived from comparisons of local and high-z SNe Ia in recent years. We describe the various aspects of the extraction in some detail as guidelines for the first-time user and present an optimal observing strategy for successful implementation of this method in future high-z SN Ia spectroscopic follow-up programmes.
△ Less
Submitted 18 October, 2004;
originally announced October 2004.
-
Spectroscopic Observations and Analysis of the Peculiar SN 1999aa
Authors:
G. Garavini,
G. Folatelli,
A. Goobar,
S. Nobili,
G. Aldering,
A. Amadon,
R. Amanullah,
P. Astier,
C. Balland,
G. Blanc,
M. S. Burns,
A. Conley,
T. Dahlen,
S. E. Deustua,
R. Ellis,
S. Fabbro,
X. Fan,
B. Frye,
E. L. Gates,
R. Gibbons,
G. Goldhaber,
B. Goldman,
D. E. Groom,
J. Haissinski,
D. Hardin
, et al. (33 additional authors not shown)
Abstract:
We present an extensive new time-series of spectroscopic data of the peculiar SN 1999aa in NGC 2595. Our data set includes 25 optical spectra between -11 and +58 days with respect to B-band maximum light, providing an unusually complete time history. The early spectra resemble those of a SN 1991T-like object but with a relatively strong Ca H&K absorption feature. The first clear sign of Si II 63…
▽ More
We present an extensive new time-series of spectroscopic data of the peculiar SN 1999aa in NGC 2595. Our data set includes 25 optical spectra between -11 and +58 days with respect to B-band maximum light, providing an unusually complete time history. The early spectra resemble those of a SN 1991T-like object but with a relatively strong Ca H&K absorption feature. The first clear sign of Si II 6355, characteristic of Type Ia supernovae, is found at day -7 and its velocity remains constant up to at least the first month after B-band maximum light. The transition to normal-looking spectra is found to occur earlier than in SN 1991T suggesting SN 1999aa as a possible link between SN 1991T-like and Branch-normal supernovae. Comparing the observations with synthetic spectra, doubly ionized Fe, Si and Ni are identified at early epochs. These are characteristic of SN 1991T-like objects. Furthermore, in the day -11 spectrum, evidence is found for an absorption feature which could be identified as high velocity C II 6580 or H-alpha. At the same epoch C III 4648.8 at photospheric velocity is probably responsible for the absorption feature at 4500 A. High velocity Ca is found around maximum light together with Si II and Fe II confined in a narrow velocity window. Implied constraints on supernovae progenitor systems and explosion hydrodynamical models are briefly discussed.
△ Less
Submitted 20 April, 2004;
originally announced April 2004.
-
New Constraints on $Ω_M$, $Ω_Λ$, and w from an Independent Set of Eleven High-Redshift Supernovae Observed with HST
Authors:
R. A. Knop,
G. Aldering,
R. Amanullah,
P. Astier,
G. Blanc,
M. S. Burns,
A. Conley,
S. E. Deustua,
M. Doi,
R. Ellis,
S. Fabbro,
G. Folatelli,
A. S. Fruchter,
G. Garavini,
S. Garmond,
K. Garton,
R. Gibbons,
G. Goldhaber,
A. Goobar,
D. E. Groom,
D. Hardin,
I. Hook,
D. A. Howell,
A. G. Kim,
B. C. Lee
, et al. (23 additional authors not shown)
Abstract:
We report measurements of $Ω_M$, $Ω_Λ$, and w from eleven supernovae at z=0.36-0.86 with high-quality lightcurves measured using WFPC-2 on the HST. This is an independent set of high-redshift supernovae that confirms previous supernova evidence for an accelerating Universe. Combined with earlier Supernova Cosmology Project data, the new supernovae yield a flat-universe measurement of the mass de…
▽ More
We report measurements of $Ω_M$, $Ω_Λ$, and w from eleven supernovae at z=0.36-0.86 with high-quality lightcurves measured using WFPC-2 on the HST. This is an independent set of high-redshift supernovae that confirms previous supernova evidence for an accelerating Universe. Combined with earlier Supernova Cosmology Project data, the new supernovae yield a flat-universe measurement of the mass density $Ω_M=0.25^{+0.07}_{-0.06}$ (statistical) $\pm0.04$ (identified systematics), or equivalently, a cosmological constant of $Ω_Λ=0.75^{+0.06}_{-0.07}$ (statistical) $\pm0.04$ (identified systematics). When the supernova results are combined with independent flat-universe measurements of $Ω_M$ from CMB and galaxy redshift distortion data, they provide a measurement of $w=-1.05^{+0.15}_{-0.20}$ (statistical) $\pm0.09$ (identified systematic), if w is assumed to be constant in time. The new data offer greatly improved color measurements of the high-redshift supernovae, and hence improved host-galaxy extinction estimates. These extinction measurements show no anomalous negative E(B-V) at high redshift. The precision of the measurements is such that it is possible to perform a host-galaxy extinction correction directly for individual supernovae without any assumptions or priors on the parent E(B-V) distribution. Our cosmological fits using full extinction corrections confirm that dark energy is required with $P(Ω_Λ>0)>0.99$, a result consistent with previous and current supernova analyses which rely upon the identification of a low-extinction subset or prior assumptions concerning the intrinsic extinction distribution.
△ Less
Submitted 12 September, 2003;
originally announced September 2003.
