Accurate photometric redshifts are a lynchpin for many future experiments to pin down the cosmological model and for studies of galaxy evolution. In this study, a novel sparse regression framework for photometric redshift estimation is presented. Sim
ulated and real data from SDSS DR12 were used to train and test the proposed models. We show that approaches which include careful data preparation and model design offer a significant improvement in comparison with several competing machine learning algorithms. Standard implementations of most regression algorithms have as the objective the minimization of the sum of squared errors. For redshift inference, however, this induces a bias in the posterior mean of the output distribution, which can be problematic. In this paper we directly target minimizing $Delta z = (z_textrm{s} - z_textrm{p})/(1+z_textrm{s})$ and address the bias problem via a distribution-based weighting scheme, incorporated as part of the optimization objective. The results are compared with other machine learning algorithms in the field such as Artificial Neural Networks (ANN), Gaussian Processes (GPs) and sparse GPs. The proposed framework reaches a mean absolute $Delta z = 0.0026(1+z_textrm{s})$, over the redshift range of $0 le z_textrm{s} le 2$ on the simulated data, and $Delta z = 0.0178(1+z_textrm{s})$ over the entire redshift range on the SDSS DR12 survey, outperforming the standard ANNz used in the literature. We also investigate how the relative size of the training set affects the photometric redshift accuracy. We find that a training set of textgreater 30 per cent of total sample size, provides little additional constraint on the photometric redshifts, and note that our GP formalism strongly outperforms ANNz in the sparse data regime for the simulated data set.
Radio continuum surveys have, in the past, been of restricted use in cosmology. Most studies have concentrated on cross-correlations with the cosmic microwave background to detect the integrated Sachs-Wolfe effect, due to the large sky areas that can
be surveyed. As we move into the SKA era, radio continuum surveys will have sufficient source density and sky area to play a major role in cosmology on the largest scales. In this chapter we summarise the experiments that can be carried out with the SKA as it is built up through the coming decade. We show that the SKA can play a unique role in constraining the non-Gaussianity parameter to sigma(f_NL) ~ 1, and provide a unique handle on the systematics that inhibit weak lensing surveys. The SKA will also provide the necessary data to test the isotropy of the Universe at redshifts of order unity and thus evaluate the robustness of the cosmological principle.Thus, SKA continuum surveys will turn radio observations into a central probe of cosmological research in the coming decades.
Radio wavelengths offer the unique possibility of tracing the total star-formation rate in galaxies, both obscured and unobscured. As such, they may provide the most robust measurement of the star-formation history of the Universe. In this chapter we
highlight the constraints that the SKA can place on the evolution of the star-formation history of the Universe, the survey area required to overcome sample variance, the spatial resolution requirements, along with the multi-wavelength ancillary data that will play a major role in maximising the scientific promise of the SKA. The required combination of depth and resolution means that a survey to trace the star formation in the Universe should be carried out with a facility that has a resolution of at least ~0.5arcsec, with high sensitivity at < 1 GHz. We also suggest a strategy that will enable new parameter space to be explored as the SKA expands over the coming decade.
Understanding the evolution of accretion activity is fundamental to our understanding of how galaxies form and evolve over the history of the Universe. We analyse a complete sample of 27 radio galaxies which includes both high-excitation (HEGs) and l
ow excitation galaxies (LEGs), spanning a narrow redshift range of 0.9 < z < 1.1 and covering a factor of ~1000 in radio luminosity. Using data from the Spitzer Space Telescope combined with ground-based optical and near-infrared imaging, we show that the host galaxies have masses in the range of 10.7 < log (M /M_sun) < 12.0 with HEGs and LEGs exhibiting no difference in their mass distributions. We also find that HEGs accrete at significantly higher rates than LEGs, with the HEG/LEG division lying at an Eddington ratio of ~0.04, which is in excellent agreement with theoretical predictions of where the accretion rate becomes radiatively inefficient, thus supporting the idea of HEGs and LEGs being powered by different modes of accretion. Our study also shows that at least up to L_151MHz ~3x10^27 W /Hz /sr, HEGs and LEGs are indistinguishable in terms of their radio properties. From this result we infer that, at least for the lower radio luminosity range, another factor besides accretion rate must play an important role in the process of triggering jet activity.
Understanding the interplay between black-hole accretion and star formation, and how to disentangle the two, is crucial to our understanding of galaxy formation and evolution. To investigate, we use a combination of optical and near-infrared photomet
ry to select a sample of 74 quasars from the VISTA Deep Extragalactic Observations (VIDEO) Survey, over 1 deg$^2$. The depth of VIDEO allows us to study very low accretion rates and/or lower-mass black holes, and 26 per cent of the candidate quasar sample has been spectroscopically confirmed. We use a radio-stacking technique to sample below the nominal flux-density threshold using data from the Very Large Array at 1.4 GHz and find, in agreement with other work, that a power-law fit to the quasar-related radio source counts is inadequate at low flux density. By comparing with a control sample of galaxies (where we match in terms of stellar mass), and by estimating the star formation rate, we suggest that this radio emission is predominantly caused by accretion activity rather than star-formation activity.
We measure star-formation rates (SFRs) and specific SFRs (SSFRs) of Ks-selected galaxies from the VIDEO survey by stacking 1.4-GHz Very Large Array data. We split the sample, which spans 0 < z < 3 and stellar masses 10**8.0 < Mstellar/Msol < 10**11.5
, into elliptical, irregular or starburst galaxies based on their spectral-energy distributions. We find that SSFR falls with stellar mass, in agreement with the `downsizing paradigm. We consider the dependence of the SSFR-mass slope on redshift: for our full and elliptical samples the slope flattens, but for the irregular and starburst samples the slope is independent of redshift. The rate of SSFR evolution reduces slightly with stellar mass for ellipticals, but irregulars and starbursts co-evolve across stellar masses. Our results for SSFR as a function of stellar mass and redshift are in agreement with those derived from other radio-stacking measurements of mass-selected passive and star-forming galaxies, but inconsistent with those generated from semi-analytic models, which tend to underestimate SFRs and SSFRs. There is a need for deeper high-resolution radio surveys such as those from telescopes like the next-generation MeerKAT in order to probe lower masses at earlier times and to permit direct detections, i.e. to study individual galaxies in detail.
We present an analysis of four complete samples of radio-loud AGN (3CRR, 2Jy, 6CE and 7CE) using near- and mid-IR data taken by the Wide-Field Infrared Survey Explorer (WISE). The combined sample consists of 79 quasars and 273 radio galaxies, and cov
ers a redshift range 0.003<z<3.395. The dichotomy in the mid-IR properties of low- and high-excitation radio galaxies (LERGs - HERGs) is analysed for the first time using large complete samples. Our results demonstrate that a division in the accretion modes of LERGs and HERGs clearly stands out in the mid-IR-radio plane (L_(22 mu m) = 5x10^(43) erg s^(-1)). This means that WISE data can be effectively used to diagnose accretion modes in radio-loud AGN. The mid-IR properties of all objects were analysed to test the unification between quasars and radio galaxies, consistent with earlier work and we argue that smooth torus models best reproduce the observation. Quasars are found to have higher mid-IR luminosities than radio galaxies. We also studied all the sources in the near-IR to gain insights into evolution of AGN host galaxies. A relation found between the near-IR luminosity and redshift, well-known in the near-IR, is apparent in the two near-IR WISE bands, supporting the idea that radio sources are hosted by massive elliptical galaxies that formed their stars at high redshifts and evolved passively thereafter. Evaluation of the positions of the sample objects in WISE colour-colour diagrams shows that widely used WISE colour cuts are not completely reliable in selecting AGN.
The shape of the curves defined by the counts of radio sources per unit area as a function of their flux density was one of the earliest cosmological probes. Radio source counts continue to be an area of interest, used to study the relative populatio
ns of galaxy types in the Universe (as well as investigate any cosmological evolution in luminosity functions). They are a vital consideration for determining how source confusion may limit the depth of a radio interferometer observation, and are essential for characterising extragalactic foregrounds in CMB experiments. There is currently no consensus as to the relative populations of the faintest (sub-mJy) source types, where the counts turn-up. Most of the source counts in this regime are gathered from multiple observations that each use a deep, single pointing with a radio interferometer. These independent measurements show large amounts of scatter (factors ~ a few) that significantly exceeds their stated uncertainties. In this article we use a simulation of the extragalactic radio continuum emission to assess the level at which sample variance may be the cause of the scatter. We find that the scatter induced by sample variance in the simulated counts decreases towards lower flux density bins as the raw source counts increase. The field-to-field variations are significant, and could even be the sole cause at >100 {mu}Jy. We present a method for evaluating the flux density limit that a survey must reach in order to reduce the count uncertainty induced by sample variance to a specific value. We also derive a method for correcting Poisson errors on counts in order to include the uncertainties due to the cosmological clustering of sources. An empirical constraint on the effect of sample variance at these low luminosities is unlikely to arise until the completion of new large-scale surveys with next-generation radio telescopes.
In this paper we describe the first data release of the the Visible and Infrared Survey Telescope for Astronomy (VISTA) Deep Extragalactic Observations (VIDEO) survey. VIDEO is a ~12degree^2 survey in the near-infrared Z,Y,J,H and K_s bands, specific
ally designed to enable the evolution of galaxies and large structures to be traced as a function of both epoch and environment from the present day out to z=4, and active galactic nuclei (AGN) and the most massive galaxies up to and into the epoch of reionization. With its depth and area, VIDEO will be able to fully explore the period in the Universe where AGN and starburst activity were at their peak and the first galaxy clusters were beginning to virialize. VIDEO therefore offers a unique data set with which to investigate the interplay between AGN, starbursts and environment, and the role of feedback at a time when it was potentially most crucial. We provide data over the VIDEO-XMM3 tile, which also covers the Canada-France-Hawaii-Telescope Legacy Survey Deep-1 field (CFHTLS-D1). The released VIDEO data reach a 5-sigma AB-magnitude depth of Z=25.7, Y=24.5, J=24.4, H=24.1 and K_s=23.8 in 2 arcsec diameter apertures (the full depth of Y=24.6 will be reached within the full integration time in future releases). The data are compared to previous surveys over this field and we find good astrometric agreement with the Two-Micron All Sky Survey, and source counts in agreement with the recently released UltraVISTA survey data. The addition of the VIDEO data to the CFHTLS-D1 optical data increases the accuracy of photometric redshifts and significantly reduces the fraction of catastrophic outliers over the redshift range 0<z<1 from 5.8 to 3.1 per cent in the absence of an i-band luminosity prior. (Truncated Abstract)
In this paper we investigate the performance of the likelihood ratio method as a tool for identifying optical and infrared counterparts to proposed radio continuum surveys with SKA precursor and pathfinder telescopes. We present a comparison of the i
nfrared counterparts identified by the likelihood ratio in the VISTA Deep Extragalactic Observations (VIDEO) survey to radio observations with 6, 10 and 15 arcsec resolution. We cross-match a deep radio catalogue consisting of radio sources with peak flux density $>$ 60 $mu$Jy with deep near-infrared data limited to $K_{mathrm{s}}lesssim$ 22.6. Comparing the infrared counterparts from this procedure to those obtained when cross-matching a set of simulated lower resolution radio catalogues indicates that degrading the resolution from 6 arcsec to 10 and 15 arcsec decreases the completeness of the cross-matched catalogue by approximately 3 and 7 percent respectively. When matching against shallower infrared data, comparable to that achieved by the VISTA Hemisphere Survey, the fraction of radio sources with reliably identified counterparts drops from $sim$89%, at $K_{mathrm{s}}lesssim$22.6, to 47% with $K_{mathrm{s}}lesssim$20.0. Decreasing the resolution at this shallower infrared limit does not result in any further decrease in the completeness produced by the likelihood ratio matching procedure. However, we note that radio continuum surveys with the MeerKAT and eventually the SKA, will require long baselines in order to ensure that the resulting maps are not limited by instrumental confusion noise.
