Photometric redshifts for the next generation of deep radio continuum surveys – I. Template fitting

Duncan, K. J.; Brown, M. J. I.; Williams, W. L.; Best, P. N.; Buat, V.; Burgarella, D.; Jarvis, Matt J.; Małek, K.; Oliver, Seb; Röttgering, H. J. A.; Smith, D. J. B.

doi:10.1093/mnras/stx2536

Cited by 77 publications

(60 citation statements)

References 94 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Including the WISE colours in the algorithm significantly narrows the ∆z distribution (σ∆z = 0.38), which is further improved with the addition of the GALEX colours (σ∆z = 0.31). Without the need for visual inspection nor the filtering out of outliers, the spread is comparable to other studies (using similar and differing methods) which include NIR photometry (Ball et al 2008;Bovy et al 2012;Brescia et al 2013;Yang et al 2017;Duncan et al 2018;Salvato et al 2019).…”

Section: Discussionsupporting

confidence: 79%

QSO photometric redshifts from SDSS, WISE, and GALEX colours

Curran

2020

Monthly Notices of the Royal Astronomical Society: Letters

View full text Add to dashboard Cite

Machine learning techniques, specifically the k-nearest neighbour algorithm applied to optical band colours, have had some success in predicting photometric redshifts of quasi-stellar objects (QSOs): Although the mean of differences between the spectroscopic and photometric redshifts, ∆z, is close to zero, the distribution of these differences remains wide and distinctly non-Gaussian. As per our previous empirical estimate of photometric redshifts, we find that the predictions can be significantly improved by adding colours from other wavebands, namely the near-infrared and ultraviolet. Self-testing this, by using half of the 33 643 strong QSO sample to train the algorithm, results in a significantly narrower spread in ∆z for the remaining half of the sample. Using the whole QSO sample to train the algorithm, the same set of magnitudes return a similar spread in ∆z for a sample of radio sources (quasars). Although the matching coincidence is relatively low (739 of the 3663 sources having photometry in the relevant bands), this is still significantly larger than from the empirical method (2%) and thus may provide a method with which to obtain redshifts for the vast number of continuum radio sources expected to be detected with the next generation of large radio telescopes.

show abstract

Section: Discussionsupporting

confidence: 79%

QSO photometric redshifts from SDSS, WISE, and GALEX colours

Curran

2020

Monthly Notices of the Royal Astronomical Society: Letters

View full text Add to dashboard Cite

show abstract

“…3) GPz: GPz is a machine learning regression algorithm originally developed for the problem in astrophysics of calculating the photometric redshifts of galaxies; the details of the algorithm and the key developments in machine learning (ML) theory are described in [6], [7], and applied to photometric redshift calculation in [19], [20], and to orbital dynamics in [21]. The algorithm is 'GP' based; a Gaussian Process is a stochastic process with a random variable defined at each point in a space of interest, such that any finite subset of the random variables has a multivariate normal distribution (equivalently any linear combination of random variables from different points has a normal distribution).…”

Section: Problem Formulation and Methodologymentioning

confidence: 99%

“…for the goal of achieving ignition at the NIF we probably have particular interest in having low RMSE close to cliff edges, but are relatively insensitive to RMSE or bias both far above and far below this boundary. We would note that CSL is a method that can in some circumstances obtain slightly better statistical properties in certain parts of parameter space [20]; but it cannot extract information from the data that simply isn't there (e.g. a really extreme weighting scheme will still fail to improve predictions in parts of parameter space with almost no data).…”

Section: ) Predictions and Optimal Designmentioning

confidence: 99%

Using Sparse Gaussian Processes for Predicting Robust Inertial Confinement Fusion Implosion Yields

Hatfield

Rose

Scott

et al. 2020

IEEE Trans. Plasma Sci.

View full text Add to dashboard Cite

Here we present the application of an advanced Sparse Gaussian Process based machine learning algorithm to the challenge of predicting the yields of inertial confinement fusion (ICF) experiments. The algorithm is used to investigate the parameter space of an extremely robust ICF design for the National Ignition Facility, the 'Simplest Design'; deuteriumtritium gas in a plastic ablator with a Gaussian, Planckian drive. In particular we show that i) GPz has the potential to decompose uncertainty on predictions into uncertainty from lack of data and shot-to-shot variation, ii) permits the incorporation of sciencegoal specific cost-sensitive learning e.g. focussing on the high-yield parts of parameter space and iii) is very fast and effective in high dimensions.1 Transfer learning is a family is of machine learning methods that seek to let an algorithm apply information/knowledge gained from one problem, to the task of solving a similar but different problem.

show abstract

“…The HELP photometry catalogue for the COSMOS field is based on the COSMOS2015 catalogue from Laigle et al (2016) (see Section 2.2.2 for a description of the COS-MOS2015 catalogue). The CIGALE catalogue we use was compiled by fitting every source within the HELP photometric catalogue for the COSMOS field that has at least four 'optical' and 'NIR' fluxes, where 'optical' bands are defined as ugrizy and N921, and 'NIR' bands are J and K (Ma lek et al 2018;Shirley et al 2019) Photometric redshifts are calculated as part of the HELP pipeline, using a Bayesian combination approach, which combines popular photometric redshift estimator templates to achieve the best estimate of the redshift, see Duncan et al (2018a) and Duncan et al (2018b). Spectroscopic redshifts are used, where possible, and are sourced from various different surveys compiled by the HELP consortium including: SDSS (Albareti et al 2017), PRIMUS (Cool et al 2013), zBRIGHT (Lilly et al 2007) and GAMA (Davies et al 2015).…”

Section: Appendix A: Cross Check Of Cosmos Source Cataloguesmentioning

confidence: 99%

S2COSMOS: Evolution of gas mass with redshift using dust emission

Millard

Eales

Smith

et al. 2020

Monthly Notices of the Royal Astronomical Society

View full text Add to dashboard Cite

We investigate the evolution of the gas mass fraction for galaxies in the COSMOS field using submillimetre emission from dust at 850µm. We use stacking methodologies on the 850 µm S2COSMOS map to derive the gas mass fraction of galaxies out to high redshifts, 0 ≤ z ≤ 5, for galaxies with stellar masses of 10 9.5 < M * (M ) < 10 11.75 . In comparison to previous literature studies we extend to higher redshifts, include more normal star-forming galaxies (on the main sequence), and also investigate the evolution of the gas mass fraction split by star-forming and passive galaxy populations. We find our stacking results broadly agree with scaling relations in the literature. We find tentative evidence for a peak in the gas mass fraction of galaxies at around z ∼ 2.5 − 3, just before the peak of the star formation history of the Universe. We find that passive galaxies are particularly devoid of gas, compared to the star-forming population. We find that even at high redshifts, high stellar mass galaxies still contain significant amounts of gas.

show abstract

Photometric redshifts for the next generation of deep radio continuum surveys – I. Template fitting

Cited by 77 publications

References 94 publications

QSO photometric redshifts from SDSS, WISE, and GALEX colours

QSO photometric redshifts from SDSS, WISE, and GALEX colours

Using Sparse Gaussian Processes for Predicting Robust Inertial Confinement Fusion Implosion Yields

S2COSMOS: Evolution of gas mass with redshift using dust emission

Contact Info

Product

Resources

About