arXiv subjects

L. Rudnick

Publications and source records attributed to L. Rudnick.

At least 55 records · Page 3Linked to original sources

arXiv2021

Downstream depolarization in the Sausage relic: a 1-4 GHz Very Large Array study

Radio relics are elongated sources related to shocks driven by galaxy cluster merger events. Although these objects are highly polarized at GHz frequencies ($\gtrsim 20\%$), high-resolution studies of their polarization properties are still lacking. We present the first high-resolution and high-sensitivity polarimetry study of the merging galaxy cluster CIZA J2242.8+5301 in the 1-4 GHz frequency band. We use the $QU$-fitting approach to model the Stokes $I$, $Q$ and $U$ emission, obtaining best-fit intrinsic polarization fraction ($p_0$), intrinsic polarization angle ($χ_0$), Rotation Measure (RM) and wavelength-dependent depolarization ($σ_{\rm RM}$) maps of the cluster. Our analysis focuses on the northern relic (RN). For the first time in a radio relic, we observe a decreasing polarization fraction in the downstream region. Our findings are possibly explained by geometrical projections and/or by decreasing of the magnetic field anisotropy towards the cluster center. From the amount of depolarization of the only detected background radio galaxy, we estimate a turbulent magnetic field strength of $B_{\rm turb}\sim5.6~μ$Gauss in the relic. Finally, we observe Rotation Measure fluctuations of about 30 rad m$^{-2}$ around at the median value of 140.8 rad m$^{-2}$ at the relic position.

astro-ph.GA↗

arXiv2021

Gamma-ray detection toward the Coma cluster with Fermi-LAT: Implications for the cosmic ray content in the hadronic scenario

The presence of relativistic electrons within the diffuse gas phase of galaxy clusters is now well established, but their detailed origin remains unclear. Cosmic ray protons are also expected to accumulate during the formation of clusters and would lead to gamma-ray emission through hadronic interactions within the thermal gas. Recently, the detection of gamma-ray emission has been reported toward the Coma cluster with Fermi-LAT. Assuming that this gamma-ray emission arises from hadronic interactions in the ICM, we aim at exploring the implication of this signal on the cosmic ray content of the Coma cluster. We use the MINOT software to build a physical model of the cluster and apply it to the Fermi-LAT data. We also consider contamination from compact sources and the impact of various systematic effects. We confirm that a significant gamma-ray signal is observed within the characteristic radius $θ_{500}$ of the Coma cluster, with a test statistic TS~27 for our baseline model. The presence of a possible point source may account for most of the observed signal. However, this source could also correspond to the peak of the diffuse emission of the cluster itself and extended models match the data better. We constrain the cosmic ray to thermal energy ratio within $R_{500}$ to $X_{\rm CRp}=1.79^{+1.11}_{-0.30}$\% and the slope of the energy spectrum of cosmic rays to $α=2.80^{+0.67}_{-0.13}$. Finally, we compute the synchrotron emission associated with the secondary electrons produced in hadronic interactions assuming steady state. This emission is about four times lower than the overall observed radio signal, so that primary cosmic ray electrons or reacceleration of secondary electrons is necessary to explain the total emission. Assuming an hadronic origin of the signal, our results provide the first quantitative measurement of the cosmic ray proton content in a cluster.[Abridged]

astro-ph.HE↗

arXiv2020

Radio observations of the merging galaxy cluster system Abell 3391-Abell 3395

The pre-merging system of galaxy clusters Abell 3391-Abell 3395 located at a mean redshift of 0.053 has been observed at 1 GHz in an ASKAP/EMU Early Science observation as well as in X-rays with eROSITA. The projected separation of the X-ray peaks of the two clusters is $\sim$50$'$ or $\sim$ 3.1 Mpc. Here we present an inventory of interesting radio sources in this field around this cluster merger. While the eROSITA observations provide clear indications of a bridge of thermal gas between the clusters, neither ASKAP nor MWA observations show any diffuse radio emission coinciding with the X-ray bridge. We derive an upper limit on the radio emissivity in the bridge region of $\langle J \rangle_{1\,{\rm GHz}}< 1.2 \times 10^{-44} {\rm W}\, {\rm Hz}^{-1} {\rm m}^{-3}$. A non-detection of diffuse radio emission in the X-ray bridge between these two clusters has implications for particle-acceleration mechanisms in cosmological large-scale structure. We also report extended or otherwise noteworthy radio sources in the 30 deg$^2$ field around Abell 3391-Abell 3395. We identified 20 Giant Radio Galaxies, plus 7 candidates, with linear projected sizes greater than 1 Mpc. The sky density of field radio galaxies with largest linear sizes of $>0.7$ Mpc is $\approx 1.7$ deg$^{-2}$, three times higher than previously reported. We find no evidence for a cosmological evolution of the population of Giant Radio Galaxies. Moreover, we find seven candidates for cluster radio relics and radio halos.

astro-ph.HE↗

arXiv2020

The Abell 3391/95 galaxy cluster system: A 15 Mpc intergalactic medium emission filament, a warm gas bridge, infalling matter clumps, and (re-) accelerated plasma discovered by combining SRG/eROSITA data with ASKAP/EMU and DECam data

We used dedicated SRG/eROSITA X-ray, ASKAP/EMU radio, and DECam optical observations of a 15 sq.deg region around the interacting galaxy cluster system A3391/95 to study the warm-hot gas in cluster outskirts and filaments, the surrounding large-scale structure and its formation process. We relate the observations to expectations from cosmological hydrodynamic simulations from the Magneticum suite. We trace the irregular morphology of warm-hot gas of the main clusters from their centers out to well beyond their characteristic radii, $r_{200}$. Between the two main cluster systems, we observe an emission bridge; thanks to eROSITA's unique soft response and large field of view, we discover tantalizing hints for warm gas. Several matter clumps physically surrounding the system are detected. For the "Northern Clump," we provide evidence that it is falling towards A3391 from the hot gas morphology and radio lobe structure of its central AGN. Many of the extended sources in the field detected by eROSITA are known clusters or new clusters in the background, including a known SZ cluster at redshift z=1. We discover an emission filament north of the virial radius, $r_{100}$, of A3391 connecting to the Northern Clump and extending south of A3395 towards another galaxy cluster. The total projected length of this continuous warm-hot emission filament is 15 Mpc, running almost 4 degrees across the entire eROSITA observation. The DECam galaxy density map shows galaxy overdensities in the same regions. The new datasets provide impressive confirmation of the theoretically expected structure formation processes on the individual system level, including the surrounding warm-hot intergalactic medium distribution compared to the Magneticum simulation. Our spatially resolved findings show that baryons indeed reside in large-scale warm-hot gas filaments with a clumpy structure.

astro-ph.CO↗

arXiv2020

Unexpected Circular Radio Objects at High Galactic Latitude

We have found a class of circular radio objects in the Evolutionary Map of the Universe Pilot Survey, using the Australian Square Kilometre Array Pathfinder telescope. The objects appear in radio images as circular edge-brightened discs, about one arcmin diameter, that are unlike other objects previously reported in the literature. We explore several possible mechanisms that might cause these objects, but none seems to be a compelling explanation.

astro-ph.GA↗

arXiv2020

Understanding the radio relic emission in the galaxy cluster MACS J0717.5+3745: spectral analysis

Radio relics are diffuse, extended synchrotron sources that originate from shock fronts generated during cluster mergers. The massive merging galaxy cluster MACS J0717.5+3745 hosts one of the more complex relics known to date. We present upgraded Giant Metrewave Radio Telescope band 3 (300-500 MHz) and band 4 (550-850 MHz) observations. These new observations, combined with published VLA and the new LOFAR HBA data, allow us to carry out a detailed, high spatial resolution spectral analysis of the relic over a broad range of frequencies. The integrated spectrum of the relic closely follows a power-law between 144 MHz and 5.5 GHz with a mean spectral slope $α=-1.16\pm0.03$. Despite its complex morphology, the subregions of the relic and the other isolated filaments also follow power-law behaviors, and show similar spectral slopes. Assuming Diffusive Shock Acceleration, we estimate a dominant Mach number of $\sim 3.7$ for the shocks that make up the relic. Comparison with recent numerical simulations suggests that in the case of radio relics, the slopes of the integrated radio spectra are determined by the Mach number of the accelerating shock, with $α$ nearly constant, namely between $-1.13$ and $-1.17$, for Mach numbers $3.5 - 4.0$. The spectral shapes inferred from spatially resolved regions show curvature, we speculate that the relic is inclined along the line-of-sight. The locus of points in the simulated color-color plots changes significantly with the relic viewing angle. We conclude that projection effects and inhomogeneities in the shock Mach number dominate the observed spectral properties of the relic in this complex system. Based on the new observations we raise the possibility that the relic and a narrow-angle-tailed radio galaxy are two different structures projected along the same line-of-sight.

astro-ph.GA↗

arXiv2020

ClG 0217+70: A massive merging galaxy cluster with a large radio halo and relics

We present an analysis of archival Chandra data of the merging galaxy cluster ClG 0217+70. The Fe XXV He$α$ X-ray emission line is clearly visible in the 25 ks observation, allowing a precise determination of the redshift of the cluster as $z=0.180\pm0.006$. We measure $kT_{500}=8.3\pm0.4$ keV and estimate $M_{500}=(1.06\pm0.11)\times10^{15}\ M_\odot$ based on existing scaling relations. Correcting both the radio and X-ray luminosities with the revised redshift reported here, which is much larger than previously inferred based on sparse optical data, this object is no longer an X-ray underluminous outlier in the $L_\mathrm{X}-P_\mathrm{radio}$ scaling relation. The new redshift also means that, in terms of physical scale, ClG 0217+70 hosts one of the largest radio halos and one of the largest radio relics known to date. Most of the relic candidates lie in projection beyond $r_{200}$. The X-ray morphological parameters suggest that the intracluster medium is still dynamically disturbed. Two X-ray surface brightness discontinuities are confirmed in the northern and southern parts of the cluster, with density jumps of $1.40\pm0.16$ and $3.0\pm0.6$, respectively. We also find a $700\times200$ kpc X-ray faint channel in the western part of the cluster, which may correspond to compressed heated gas or increased non-thermal pressure due to turbulence or magnetic fields.

astro-ph.HE↗

arXiv2020

Radio Galaxy Zoo: New Giant Radio Galaxies in the RGZ DR1catalogue

In this paper, we present the identification of five previously unknown giant radio galaxies (GRGs) using Data Release 1 of the Radio Galaxy Zoo citizen science project and a selection method appropriate to the training and validation of deep learning algorithms for new radio surveys. We associate one of these new GRGs with the brightest cluster galaxy (BCG) in the galaxy cluster GMBCG J251.67741+36.45295 and use literature data to identify a further 13 previously known GRGs as BCG candidates, increasing the number of known BCG GRGs by >60%. By examining local galaxy number densities for the number of all known BCG GRGs, we suggest that the existence of this growing number implies that GRGs are able to reside in the centers of rich ($\sim 10^{14}$ M$_{\odot}$) galaxy clusters and challenges the hypothesis that GRGs grow to such sizes only in locally under-dense environments.

astro-ph.GA↗

arXiv2020

New mysteries and challenges from the Toothbrush relic: wideband observations from 550 MHz to 8 GHz

Radio relics are diffuse extended synchrotron sources that originate from shock fronts induced by galaxy cluster mergers. The actual particle acceleration mechanism at the shock fronts is still under debate. The galaxy cluster 1RXS J0603.3+4214 hosts one of the most intriguing examples of radio relics, known as the Toothbrush. We present new wideband radio continuum observations made with uGMRT and VLA. Our new observations, in combination with previously published data, allowed us to carry out a detailed high spatial resolution spectral and curvature analysis of the known diffuse radio emission sources, over a broad range of frequencies. The integrated spectrum of the Toothbrush follows closely a power law over close to 2 decades in frequency, with a spectral index of $-1.16\pm0.02$. We do not find any evidence of spectral steepening below 8 GHz. The subregions of the Toothbrush show an identical spectral slopes, suggesting that observed spectral index is rather set by the distribution of Mach numbers which may have a similar shape at different parts of the shock front. Indeed, numerical simulations show an intriguing similar spectral index, indicating that the radio spectrum is dominated by the average over the inhomogeneities within the shock, with most of the emission coming from the tail of the Mach number distribution. In contrast to the Toothbrush, the spectrum of the fainter relics show a high frequency steepening. The integrated spectrum of the halo follows a power law from 150 MHz to 3 GHz with a spectral index of $-1.16\pm0.04$. We do not find any evidence for spectral curvature, not even in subareas of the halo. This suggest a homogeneous acceleration throughout the cluster volume. Between the brush region of the Toothbrush and the halo, the color-color analysis revealed emission that was consistent with an overlap between the two different spectral regions.

astro-ph.HE↗

arXiv2020

PKS 2250$-$351: A Giant Radio Galaxy in Abell 3936

We present a detailed analysis of the radio galaxy PKS 2250-351, a giant of 1.2 Mpc projected size, its host galaxy, and its environment. We use radio data from the Murchison Widefield Array, the upgraded Giant Metre-wavelength Radio Telescope, the Australian Square Kilometre Array Pathfinder, and the Australia Telescope Compact Array to model the jet power and age. Optical and infra-red data come from the Galaxy And Mass Assembly (GAMA) survey and provide information on the host galaxy and environment. GAMA spectroscopy confirms that PKS 2250-351 lies at z=0.2115 in the irregular, and likely unrelaxed, cluster Abell 3936. We find its host is a massive, `red and dead' elliptical galaxy with negligible star formation but with a highly obscured active galactic nucleus dominating the mid-infrared emission. Assuming it lies on the local M-sigma relation it has an Eddington accretion rate of lambda_EDD~0.014. We find that the lobe-derived jet power (a time-averaged measure) is an order of magnitude greater than the hotspot-derived jet power (an instantaneous measure). We propose that over the lifetime of the observed radio emission (~300 Myr) the accretion has switched from an inefficient advection dominated mode to a thin-disc efficient mode, consistent with the decrease in jet power. We also suggest that the asymmetric radio morphology is due to its environment, with the host of PKS 2250-351 lying to the west of the densest concentration of galaxies in Abell 3936.

astro-ph.GA↗

arXiv2019

The Karl G. Jansky Very Large Array Sky Survey (VLASS). Science case and survey design

The Very Large Array Sky Survey (VLASS) is a synoptic, all-sky radio sky survey with a unique combination of high angular resolution ($\approx$2.5"), sensitivity (a 1$σ$ goal of 70 $μ$Jy/beam in the coadded data), full linear Stokes polarimetry, time domain coverage, and wide bandwidth (2-4 GHz). The first observations began in September 2017, and observing for the survey will finish in 2024. VLASS will use approximately 5500 hours of time on the Karl G. Jansky Very Large Array (VLA) to cover the whole sky visible to the VLA (Declination $>-40^{\circ}$), a total of 33,885 deg$^2$. The data will be taken in three epochs to allow the discovery of variable and transient radio sources. The survey is designed to engage radio astronomy experts, multi-wavelength astronomers, and citizen scientists alike. By utilizing an "on the fly" interferometry mode, the observing overheads are much reduced compared to a conventional pointed survey. In this paper, we present the science case and observational strategy for the survey, and also results from early survey observations.

astro-ph.IM↗

arXiv2019

The ASKAP-EMU Early Science Project:Radio Continuum Survey of the Small Magellanic Cloud

We present two new radio continuum images from the Australian Square Kilometre Array Pathfinder (ASKAP) survey in the direction of the Small Magellanic Cloud (SMC). These images are part of the Evolutionary Map of the Universe (EMU) Early Science Project (ESP) survey of the Small and Large Magellanic Clouds. The two new source lists produced from these images contain radio continuum sources observed at 960 MHz (4489 sources) and 1320 MHz (5954 sources) with a bandwidth of 192 MHz and beam sizes of 30.0"x30.0" and 16.3"x15.1", respectively. The median Root Mean Squared (RMS) noise values are 186$μ$Jy beam$^{-1}$ (960 MHz) and 165$μ$Jy beam$^{-1}$ (1320 MHz). To create point source catalogues, we use these two source lists, together with the previously published Molonglo Observatory Synthesis Telescope (MOST) and the Australia Telescope Compact Array (ATCA) point source catalogues to estimate spectral indices for the whole population of radio point sources found in the survey region. Combining our ASKAP catalogues with these radio continuum surveys, we found 7736 point-like sources in common over an area of 30 deg$^2$. In addition, we report the detection of two new, low surface brightness supernova remnant candidates in the SMC. The high sensitivity of the new ASKAP ESP survey also enabled us to detect the bright end of the SMC planetary nebula sample, with 22 out of 102 optically known planetary nebulae showing point-like radio continuum emission. Lastly, we present several morphologically interesting background radio galaxies.

astro-ph.GA↗

arXiv2019

Differences in Faraday Rotation Between Adjacent Extragalactic Radio Sources as a Probe of Cosmic Magnetic Fields

Faraday rotation measures (RMs) of extragalactic radio sources provide information on line-of-sight magnetic fields, including contributions from our Galaxy, source environments, and the intergalactic medium (IGM). Looking at differences in RMs, $Δ$RM, between adjacent sources on the sky can help isolate these different components. In this work, we classify adjacent polarized sources in the NVSS as random or physical pairs. We recompute and correct the uncertainties in the NVSS RM catalog, since these were significantly overestimated. Our sample contains 317 physical and 5111 random pairs, all with Galactic latitudes $|b|\ge20^{\circ}$, polarization fractions $\ge2\%$, and angular separations between $1.^{'},$ and $20^{'}$. We find an rms $Δ$RM of $14.9\pm0.4\,$rad m$^{-2}$ and $4.6\pm1.1\,$rad m$^{-2}$ for random and physical pairs, respectively. This means polarized extragalactic sources that are close on the sky, but at different redshifts, have larger differences in RM than two components of one source. This difference of $\sim10\,$rad m$^{-2}$ is significant at $5σ$, and persists in different data subsamples. While there have been other statistical studies of $Δ$RM between adjacent polarized sources, this is the first unambiguous demonstration that some of this RM difference must be extragalactic, thereby providing a firm upper limit on the RM contribution of the IGM. If the $Δ$RMs originate local to the sources, then the local magnetic field difference between random sources is a factor of two larger than between components of one source. Alternatively, attributing the difference in $Δ$RMs to the intervening IGM yields an upper limit on the IGM magnetic field strength of $40\,$nG.

astro-ph.CO↗

arXiv2019

Radio Galaxy Zoo: Knowledge Transfer Using Rotationally Invariant Self-Organising Maps

With the advent of large scale surveys the manual analysis and classification of individual radio source morphologies is rendered impossible as existing approaches do not scale. The analysis of complex morphological features in the spatial domain is a particularly important task. Here we discuss the challenges of transferring crowdsourced labels obtained from the Radio Galaxy Zoo project and introduce a proper transfer mechanism via quantile random forest regression. By using parallelized rotation and flipping invariant Kohonen-maps, image cubes of Radio Galaxy Zoo selected galaxies formed from the FIRST radio continuum and WISE infrared all sky surveys are first projected down to a two-dimensional embedding in an unsupervised way. This embedding can be seen as a discretised space of shapes with the coordinates reflecting morphological features as expressed by the automatically derived prototypes. We find that these prototypes have reconstructed physically meaningful processes across two channel images at radio and infrared wavelengths in an unsupervised manner. In the second step, images are compared with those prototypes to create a heat-map, which is the morphological fingerprint of each object and the basis for transferring the user generated labels. These heat-maps have reduced the feature space by a factor of 248 and are able to be used as the basis for subsequent ML methods. Using an ensemble of decision trees we achieve upwards of 85.7% and 80.7% accuracy when predicting the number of components and peaks in an image, respectively, using these heat-maps. We also question the currently used discrete classification schema and introduce a continuous scale that better reflects the uncertainty in transition between two classes, caused by sensitivity and resolution limits.

astro-ph.IM↗

arXiv2018

Radio Galaxy Zoo: Observational evidence for environment as the cause of radio source asymmetry

We investigate the role of environment on radio galaxy properties by constructing a sample of large ($\gtrsim100$~kpc), nearby ($z<0.3$) radio sources identified as part of the Radio Galaxy Zoo citizen science project. Our sample consists of 16 Fanaroff-Riley Type II (FR-II) sources, 6 FR-I sources, and one source with a hybrid morphology. FR-I sources appear to be hosted by more massive galaxies, consistent with previous studies. In the FR-II sample, we compare the degree of asymmetry in radio lobe properties to asymmetry in the radio source environment, quantified through optical galaxy clustering. We find that the length of radio lobes in FR-II sources is anti-correlated with both galaxy clustering and lobe luminosity. These results are in quantitative agreement with predictions from radio source dynamical models, and suggest that galaxy clustering provides a useful proxy for the ambient gas density distribution encountered by the radio lobes.

astro-ph.GA↗

arXiv2018

Deep Very Large Array observations of the merging cluster CIZA J2242.8+5301: continuum and spectral imaging

Despite progress in understanding radio relics, there are still open questions regarding the underlying particle acceleration mechanisms. In this paper we present deep 1--4 GHz VLA observations of CIZA\,J2242.8+5301 ($z=0.1921$), a double radio relic cluster characterized by small projection on the plane of the sky. Our VLA observations reveal, for the first time, the complex morphology of the diffuse sources and the filamentary structure of the northern relic. We discover new faint diffuse radio emission extending north of the main northern relic. Our Mach number estimates for the northern and southern relics, based on the radio spectral index map obtained using the VLA observations and existing LOFAR and GMRT data, are consistent with previous radio and X-ray studies ($\mathcal{M}_{\rm RN}=2.58\pm0.17$ and $\mathcal{M}_{\rm RS}=2.10\pm0.08$). However, color-color diagrams and modelings suggest a flatter injection spectral index than the one obtained from the spectral index map, indicating that projection effects might be not entirely negligible. The southern relic consists of five "arms". Embedded in it, we find a tailed radio galaxy which seems to be connected to the relic. A spectral index flattening, where the radio tail connects to the relic, is also measured. We propose that the southern relic may trace AGN fossil electrons that are re-accelerated at a shock, with an estimated strength of $\mathcal{M}=2.4$. High-resolution mapping of other tailed radio galaxies also supports a scenario where AGN fossil electrons are revived by the merger event and could be related to the formation of some diffuse cluster radio emission.

astro-ph.GA↗

arXiv2018

Radio Galaxy Zoo: Machine learning for radio source host galaxy cross-identification

We consider the problem of determining the host galaxies of radio sources by cross-identification. This has traditionally been done manually, which will be intractable for wide-area radio surveys like the Evolutionary Map of the Universe (EMU). Automated cross-identification will be critical for these future surveys, and machine learning may provide the tools to develop such methods. We apply a standard approach from computer vision to cross-identification, introducing one possible way of automating this problem, and explore the pros and cons of this approach. We apply our method to the 1.4 GHz Australian Telescope Large Area Survey (ATLAS) observations of the Chandra Deep Field South (CDFS) and the ESO Large Area ISO Survey South 1 (ELAIS-S1) fields by cross-identifying them with the Spitzer Wide-area Infrared Extragalactic (SWIRE) survey. We train our method with two sets of data: expert cross-identifications of CDFS from the initial ATLAS data release and crowdsourced cross-identifications of CDFS from Radio Galaxy Zoo. We found that a simple strategy of cross-identifying a radio component with the nearest galaxy performs comparably to our more complex methods, though our estimated best-case performance is near 100 per cent. ATLAS contains 87 complex radio sources that have been cross-identified by experts, so there are not enough complex examples to learn how to cross-identify them accurately. Much larger datasets are therefore required for training methods like ours. We also show that training our method on Radio Galaxy Zoo cross-identifications gives comparable results to training on expert cross-identifications, demonstrating the value of crowdsourced training data.

astro-ph.IM↗

arXiv2018

Radio Galaxy Zoo: Compact and extended radio source classification with deep learning

Machine learning techniques have been increasingly useful in astronomical applications over the last few years, for example in the morphological classification of galaxies. Convolutional neural networks have proven to be highly effective in classifying objects in image data. The current work aims to establish when multiple components are present, in the astronomical context of synthesis imaging observations of radio sources. To this effect, we design a convolutional neural network to differentiate between different morphology classes using sources from the Radio Galaxy Zoo (RGZ) citizen science project. In this first step, we focus on exploring the factors that affect the performance of such neural networks, such as the amount of training data, number and nature of layers and the hyperparameters. We begin with a simple experiment in which we only differentiate between two extreme morphologies, using compact and multiple component extended sources. We found that a three convolutional layer architecture yielded very good results, achieving a classification accuracy of 97.4% on a test data set. The same architecture was then tested on a four-class problem where we let the network classify sources into compact and three classes of extended sources, achieving a test achieving a test accuracy of 93.5%. The best-performing convolutional neural network setup has been verified against RGZ Data Release 1 where a final test accuracy of 94.8% was obtained, using both original and augmented images. The use of sigma clipping does not offer a significant benefit overall, except in cases with a small number of training images.

astro-ph.IM↗