scholarly journals The PAU Survey: Photometric redshifts using transfer learning from simulations

2020 ◽  
Vol 497 (4) ◽  
pp. 4565-4579
Author(s):  
M Eriksen ◽  
A Alarcon ◽  
L Cabayol ◽  
J Carretero ◽  
R Casas ◽  
...  
Keyword(s):  
Transfer Learning ◽  
Broad Band ◽  
Training Data ◽  
Machine Learning Techniques ◽  
Test Case ◽  
Photometric Redshifts ◽  
Mixture Density ◽  
Photometric Redshift ◽  
The Galaxy ◽  
The Impact

ABSTRACT In this paper, we introduce the deepz deep learning photometric redshift (photo-z) code. As a test case, we apply the code to the PAU survey (PAUS) data in the COSMOS field. deepz reduces the σ68 scatter statistic by 50 per cent at iAB = 22.5 compared to existing algorithms. This improvement is achieved through various methods, including transfer learning from simulations where the training set consists of simulations as well as observations, which reduces the need for training data. The redshift probability distribution is estimated with a mixture density network (MDN), which produces accurate redshift distributions. Our code includes an autoencoder to reduce noise and extract features from the galaxy SEDs. It also benefits from combining multiple networks, which lowers the photo-z scatter by 10 per cent. Furthermore, training with randomly constructed coadded fluxes adds information about individual exposures, reducing the impact of photometric outliers. In addition to opening up the route for higher redshift precision with narrow bands, these machine learning techniques can also be valuable for broad-band surveys.

scholarly journals Impact of photometric redshifts on the galaxy power spectrum and BAO scale in the LSST survey

2019 ◽  
Vol 623 ◽  
pp. A76 ◽  
Author(s):  
Reza Ansari ◽  
Adeline Choyer ◽  
Farhang Habibi ◽  
Christophe Magneville ◽  
Marc Moniez ◽  
...  
Keyword(s):  
Power Spectrum ◽  
Large Scale ◽  
Near Infrared ◽  
Linear Part ◽  
Cosmological Parameters ◽  
Photometric Redshifts ◽  
Photometric Redshift ◽  
Near Ultraviolet ◽  
The Galaxy ◽  
The Impact

Context. The Large Synoptic Survey Telescope (LSST) survey will image billions of galaxies every few nights for ten years, and as such, should be a major contributor to precision cosmology in the 2020s. High precision photometric data will be available in six bands, from near-infrared to near-ultraviolet. The computation of precise, unbiased, photometric redshifts up to at least z = 2 is one of the main LSST challenges and its performance will have major impact on all extragalactic LSST sciences. Aims. We evaluate the efficiency of our photometric redshift reconstruction on mock galaxy catalogues up to z = 2.45 and estimate the impact of realistic photometric redshift (photo-z) reconstruction on the large-scale structures (LSS) power spectrum and the baryonic acoustic oscillation (BAO) scale determination for a LSST-like photometric survey. We study the effectiveness of the BAO scale as a cosmological probe in the LSST survey. Methods. We have performed a detailed modelling of the photo-z distribution as a function of galaxy type, redshift and absolute magnitude using our photo-z reconstruction code with a quality selection cut based on a boosted decision tree (BDT). We have simulated a catalogue of galaxies in the redshift range [0.2−2.45] using the Planck 2015 ΛCDM cosmological parameters over 10 000 square-degrees, in the six bands, assuming LSST photometric precision for a ten-year survey. The mock galaxy catalogues were produced with several redshift error models. The LSS power spectrum was then computed in several redshift ranges and for each error model. Finally we extracted the BAO scale and its uncertainty using only the linear part of the LSS spectrum. Results. We have computed the fractional error on the recovered power spectrum which is dominated by the shot noise at high redshift (z ≳ 1), for scales k ≳ 0.1, due to the photo-z damping. The BAO scale can be recovered with a percent or better accuracy level from z = 0.5 to z = 1.5 using realistic photo-z reconstruction. Conclusions. Reaching the LSST requirements for photo-z reconstruction is crucial to exploit the LSST potential in cosmology, in particular to measure the LSS power spectrum and its evolution with redshift. Although the BAO scale is not the most powerful cosmological probe in LSST, it can be used to check the consistency of the LSS measurement. Moreover we show that the impact of photo-z smearing on the recovered isotropic BAO scale in LSST should stay limited up to z ≈ 1.5, so as long as the galaxy number density balances the photo-z smoothing.

scholarly journals Copy-Move Forgery Detection (CMFD) Using Deep Learning for Image and Video Forensics

2021 ◽  
Vol 7 (3) ◽  
pp. 59
Author(s):  
Yohanna Rodriguez-Ortega ◽  
Dora M. Ballesteros ◽  
Diego Renza
Keyword(s):  
Deep Learning ◽  
Transfer Learning ◽  
Training Data ◽  
Video Editing ◽  
Forgery Detection ◽  
Copy Move Forgery Detection ◽  
The Impact ◽  
And Training ◽  
Selection Of ◽  
Traditional Image

With the exponential growth of high-quality fake images in social networks and media, it is necessary to develop recognition algorithms for this type of content. One of the most common types of image and video editing consists of duplicating areas of the image, known as the copy-move technique. Traditional image processing approaches manually look for patterns related to the duplicated content, limiting their use in mass data classification. In contrast, approaches based on deep learning have shown better performance and promising results, but they present generalization problems with a high dependence on training data and the need for appropriate selection of hyperparameters. To overcome this, we propose two approaches that use deep learning, a model by a custom architecture and a model by transfer learning. In each case, the impact of the depth of the network is analyzed in terms of precision (P), recall (R) and F1 score. Additionally, the problem of generalization is addressed with images from eight different open access datasets. Finally, the models are compared in terms of evaluation metrics, and training and inference times. The model by transfer learning of VGG-16 achieves metrics about 10% higher than the model by a custom architecture, however, it requires approximately twice as much inference time as the latter.

scholarly journals Simulating JWST deep extragalactic imaging surveys and physical parameter recovery

2020 ◽  
Vol 640 ◽  
pp. A67
Author(s):  
O. B. Kauffmann ◽  
O. Le Fèvre ◽  
O. Ilbert ◽  
J. Chevallard ◽  
C. C. Williams ◽  
...  
Keyword(s):  
Star Formation ◽  
Near Infrared ◽  
High Redshift ◽  
Spectral Energy Distribution ◽  
Formation Rate ◽  
Spectral Energy ◽  
Photometric Redshifts ◽  
The Galaxy ◽  
Stellar Masses ◽  
The Impact

We present a new prospective analysis of deep multi-band imaging with the James Webb Space Telescope (JWST). In this work, we investigate the recovery of high-redshift 5 <  z <  12 galaxies through extensive image simulations of accepted JWST programs, including the Early Release Science in the EGS field and the Guaranteed Time Observations in the HUDF. We introduced complete samples of ∼300 000 galaxies with stellar masses of log(M*/M⊙) > 6 and redshifts of 0 <  z <  15, as well as galactic stars, into realistic mock NIRCam, MIRI, and HST images to properly describe the impact of source blending. We extracted the photometry of the detected sources, as in real images, and estimated the physical properties of galaxies through spectral energy distribution fitting. We find that the photometric redshifts are primarily limited by the availability of blue-band and near-infrared medium-band imaging. The stellar masses and star formation rates are recovered within 0.25 and 0.3 dex, respectively, for galaxies with accurate photometric redshifts. Brown dwarfs contaminating the z >  5 galaxy samples can be reduced to < 0.01 arcmin−2 with a limited impact on galaxy completeness. We investigate multiple high-redshift galaxy selection techniques and find that the best compromise between completeness and purity at 5 <  z <  10 using the full redshift posterior probability distributions. In the EGS field, the galaxy completeness remains higher than 50% at magnitudes mUV <  27.5 and at all redshifts, and the purity is maintained above 80 and 60% at z ≤ 7 and 10, respectively. The faint-end slope of the galaxy UV luminosity function is recovered with a precision of 0.1–0.25, and the cosmic star formation rate density within 0.1 dex. We argue in favor of additional observing programs covering larger areas to better constrain the bright end.

scholarly journals Quantified uncertainties in fission yields from machine learning

2020 ◽  
Vol 242 ◽  
pp. 05003
Author(s):  
A.E. Lovell ◽  
A.T. Mohan ◽  
P. Talou ◽  
M. Chertkov
Keyword(s):  
Machine Learning ◽  
Nuclear Physics ◽  
Experimental Error ◽  
Training Data ◽  
Learning Methods ◽  
Mixture Density ◽  
Machine Learning Methods ◽  
Predicted Values ◽  
The Impact ◽  
Fission Yields

As machine learning methods gain traction in the nuclear physics community, especially those methods that aim to propagate uncertainties to unmeasured quantities, it is important to understand how the uncertainty in the training data coming either from theory or experiment propagates to the uncertainty in the predicted values. Gaussian Processes and Bayesian Neural Networks are being more and more widely used, in particular to extrapolate beyond measured data. However, studies are typically not performed on the impact of the experimental errors on these extrapolated values. In this work, we focus on understanding how uncertainties propagate from input to prediction when using machine learning methods. We use a Mixture Density Network (MDN) to incorporate experimental error into the training of the network and construct uncertainties for the associated predicted quantities. Systematically, we study the effect of the size of the experimental error, both on the reproduced training data and extrapolated predictions for fission yields of actinides.

scholarly journals The XXL Survey

2018 ◽  
Vol 620 ◽  
pp. A13 ◽  
Author(s):  
M. Ricci ◽  
C. Benoist ◽  
S. Maurogordato ◽  
C. Adami ◽  
L. Chiappetti ◽  
...  
Keyword(s):  
Near Infrared ◽  
Mass Range ◽  
Selection Function ◽  
Photometric Redshifts ◽  
Statistical Tool ◽  
Cluster Galaxy ◽  
Photometric Redshift ◽  
Luminosity Functions ◽  
The Galaxy ◽  
The Individual

Context. The luminosity function (LF) is a powerful statistical tool used to describe galaxies and learn about their evolution. In particular, the LFs of galaxies inside clusters allow us to better understand how galaxies evolve in these dense environments. Knowledge of the LFs of galaxies in clusters is also crucial for clusters studies in the optical and near-infrared (NIR) as they encode, along with their density profiles, most of their observational properties. However, no consensus has been reached yet about the evolution of the cluster galaxy LF with halo mass and redshift. Aims. The main goal of this study is to investigate the LF of a sample of 142 X-ray selected clusters, with spectroscopic redshift confirmation and a well defined selection function, spanning a wide redshift and mass range, and to test the LF dependence on cluster global properties, in a homogeneous and unbiased way. Methods. Our study is based on the Canada–France–Hawaii Telescope Legacy Survey (CFHTLS) photometric galaxy catalogue, associated with photometric redshifts. We constructed LFs inside a scaled radius using a selection in photometric redshift around the cluster spectroscopic redshift in order to reduce projection effects. The width of the photometric redshift selection was carefully determined to avoid biasing the LF and depended on both the cluster redshift and the galaxy magnitudes. The purity was then enhanced by applying a precise background subtraction. We constructed composite luminosity functions (CLFs) by stacking the individual LFs and studied their evolution with redshift and richness, analysing separately the brightest cluster galaxy (BCG) and non-BCG members. We fitted the dependences of the CLFs and BCG distributions parameters with redshift and richness conjointly in order to distinguish between these two effects. Results. We find that the usual photometric redshift selection methods can bias the LF estimate if the redshift and magnitude dependence of the photometric redshift quality is not taken into account. Our main findings concerning the evolution of the galaxy luminosity distribution with redshift and richness are that, in the inner region of clusters and in the redshift-mass range we probe (about 0 < z < 1 and 1013 M⊙ < M500 < 5 × 1014 M⊙), the bright part of the LF (BCG excluded) does not depend much on mass or redshift except for its amplitude, whereas the BCG luminosity increases both with redshift and richness.

scholarly journals PS1-STRM: neural network source classification and photometric redshift catalogue for PS1 3π DR1

2020 ◽  
Vol 500 (2) ◽  
pp. 1633-1644
Author(s):  
Róbert Beck ◽  
István Szapudi ◽  
Heather Flewelling ◽  
Conrad Holmberg ◽  
Eugene Magnier ◽  
...  
Keyword(s):  
Neural Network ◽  
Broad Band ◽  
Monte Carlo Sampling ◽  
Training Data ◽  
Space Telescopes ◽  
Absolute Deviation ◽  
Self Organizing Maps ◽  
Validation Data ◽  
Data Set ◽  
Photometric Redshift

ABSTRACT The Pan-STARRS1 (PS1) 3π survey is a comprehensive optical imaging survey of three quarters of the sky in the grizy broad-band photometric filters. We present the methodology used in assembling the source classification and photometric redshift (photo-z) catalogue for PS1 3π Data Release 1, titled Pan-STARRS1 Source Types and Redshifts with Machine learning (PS1-STRM). For both main data products, we use neural network architectures, trained on a compilation of public spectroscopic measurements that has been cross-matched with PS1 sources. We quantify the parameter space coverage of our training data set, and flag extrapolation using self-organizing maps. We perform a Monte Carlo sampling of the photometry to estimate photo-z uncertainty. The final catalogue contains 2902 054 648 objects. On our validation data set, for non-extrapolated sources, we achieve an overall classification accuracy of $98.1{{\ \rm per\ cent}}$ for galaxies, $97.8{{\ \rm per\ cent}}$ for stars, and $96.6{{\ \rm per\ cent}}$ for quasars. Regarding the galaxy photo-z estimation, we attain an overall bias of 〈Δznorm〉 = 0.0005, a standard deviation of σ(Δznorm) = 0.0322, a median absolute deviation of MAD(Δznorm) = 0.0161, and an outlier fraction of $P\left(|\Delta z_{\mathrm{norm}}|\gt 0.15\right)=1.89{{\ \rm per\ cent}}$. The catalogue will be made available as a high-level science product via the Mikulski Archive for Space Telescopes.

scholarly journals The Close AGN Reference Survey (CARS)

2019 ◽  
Vol 627 ◽  
pp. A53 ◽  
Author(s):  
B. Husemann ◽  
J. Scharwächter ◽  
T. A. Davis ◽  
M. Pérez-Torres ◽  
I. Smirnova-Pinchukova ◽  
...  
Keyword(s):  
Star Formation ◽  
Near Infrared ◽  
Formation Rate ◽  
Broad Band ◽  
Host Galaxy ◽  
Radio Jets ◽  
Spatially Resolved ◽  
The Galaxy ◽  
Multi Phase ◽  
The Impact

Context. Galaxy-wide outflows driven by star formation and/or an active galactic nucleus (AGN) are thought to play a crucial rule in the evolution of galaxies and the metal enrichment of the inter-galactic medium. Direct measurements of these processes are still scarce and new observations are needed to reveal the nature of outflows in the majority of the galaxy population. Aims. We combine extensive, spatially-resolved, multi-wavelength observations, taken as part of the Close AGN Reference Survey (CARS), for the edge-on disc galaxy HE 1353−1917 in order to characterise the impact of the AGN on its host galaxy via outflows and radiation. Methods. Multi-color broad-band photometry was combined with spatially-resolved optical, near-infrared (NIR) and sub-mm and radio observations taken with the Multi-Unit Spectroscopy Explorer (MUSE), the Near-infrared Integral Field Spectrometer (NIFS), the Atacama Large Millimeter Array (ALMA), and the Karl G. Jansky Very Large Array (VLA) to map the physical properties and kinematics of the multi-phase interstellar medium. Results. We detect a biconical extended narrow-line region ionised by the luminous AGN orientated nearly parallel to the galaxy disc, extending out to at least 25 kpc. The extra-planar gas originates from galactic fountains initiated by star formation processes in the disc, rather than an AGN outflow, as shown by the kinematics and the metallicity of the gas. Nevertheless, a fast, multi-phase, AGN-driven outflow with speeds up to 1000 km s−1 is detected close to the nucleus at 1 kpc distance. A radio jet, in connection with the AGN radiation field, is likely responsible for driving the outflow as confirmed by the energetics and the spatial alignment of the jet and multi-phase outflow. Evidence for negative AGN feedback suppressing the star formation rate (SFR) is mild and restricted to the central kpc. But while any SFR suppression must have happened recently, the outflow has the potential to greatly impact the future evolution of the galaxy disc due to its geometrical orientation. Conclusions.. Our observations reveal that low-power radio jets can play a major role in driving fast, multi-phase, galaxy-scale outflows even in radio-quiet AGN. Since the outflow energetics for HE 1353−1917 are consistent with literature, scaling relation of AGN-driven outflows the contribution of radio jets as the driving mechanisms still needs to be systematically explored.

scholarly journals The Impact of Stochastic Attenuation on Photometric Redshift Estimates

2006 ◽  
Vol 2 (S235) ◽  
pp. 438-439
Author(s):  
Thorsten Tepper García ◽  
Uta Fritze-von Alvensleben
Keyword(s):  
Star Formation ◽  
Star Formation Rate ◽  
Formation Rate ◽  
Spectral Energy ◽  
Energy Distributions ◽  
Photometric Redshifts ◽  
Lyman Alpha ◽  
Photometric Redshift ◽  
The Impact ◽  
Johnson System

AbstractWe model the stochastic attenuation by HI absorbers in the intergalactic medium (IGM), such as Lyα Forest clouds, and absorbers associated with galaxies, such as Lyman Limit systems (LLS) and Damped Lyman Alpha absorbers (DLAs), and compute an ensemble of 4 · 103 attenuated Spectral Energy Distributions (SEDs) in the Johnson system for the spectrum of a galaxy with a constant star formation rate (CSFR). Using these, we asses the impact of the stochastic attenuation on the estimates of photometric redshifts for this type of galaxy by comparison with model SEDs that include only a mean attenuation.

scholarly journals The PAU Survey: narrow-band photometric redshifts using Gaussian processes

2021 ◽  
Vol 503 (3) ◽  
pp. 4118-4135
Author(s):  
John Y H Soo ◽  
Benjamin Joachimi ◽  
Martin Eriksen ◽  
Małgorzata Siudek ◽  
Alex Alarcon ◽  
...  
Keyword(s):  
Gaussian Processes ◽  
Performance Metrics ◽  
Narrow Band ◽  
Broad Band ◽  
Reference Sample ◽  
Apparent Size ◽  
Distribution Functions ◽  
Accelerating Universe ◽  
Photometric Redshift ◽  
The Galaxy

ABSTRACT We study the performance of the hybrid template machine learning photometric redshift (photo-z) algorithm delight, which uses Gaussian processes, on a subset of the early data release of the Physics of the Accelerating Universe Survey (PAUS). We calibrate the fluxes of the 40 PAUS narrow bands with six broad-band fluxes (uBVriz) in the Cosmic Evolution Survey (COSMOS) field using three different methods, including a new method that utilizes the correlation between the apparent size and overall flux of the galaxy. We use a rich set of empirically derived galaxy spectral templates as guides to train the Gaussian process, and we show that our results are competitive with other standard photometric redshift algorithms. delight achieves a photo-z 68th percentile error of σ68 = 0.0081(1 + z) without any quality cut for galaxies with iauto &lt; 22.5 as compared to 0.0089(1 + z) and 0.0202(1 + z) for the bpz and annz2 codes, respectively. delight is also shown to produce more accurate probability distribution functions for individual redshift estimates than bpz and annz2. Common photo-z outliers of delight and bcnz2 (previously applied to PAUS) are found to be primarily caused by outliers in the narrow-band fluxes, with a small number of cases potentially indicating spectroscopic redshift failures in the reference sample. In the process, we introduce performance metrics derived from the results of bcnz2 and delight, allowing us to achieve a photo-z quality of σ68 &lt; 0.0035(1 + z) at a magnitude of iauto &lt; 22.5 while keeping 50 per cent objects of the galaxy sample.

scholarly journals Virtual to Real-World Transfer Learning: A Systematic Review

Electronics ◽  
2021 ◽  
Vol 10 (12) ◽  
pp. 1491
Author(s):  
Mahesh Ranaweera ◽  
Qusay H. Mahmoud
Keyword(s):  
Machine Learning ◽  
Systematic Review ◽  
Transfer Learning ◽  
Real World ◽  
High Performance ◽  
Research Area ◽  
Training Data ◽  
Machine Learning Techniques ◽  
Current Status ◽  
The Real

Machine learning has become an important research area in many domains and real-world applications. The prevailing assumption in traditional machine learning techniques, that training and testing data should be of the same domain, is a challenge. In the real world, gathering enough training data to create high-performance learning models is not easy. Sometimes data are not available, very expensive, or dangerous to collect. In this scenario, the concept of machine learning does not hold up to its potential. Transfer learning has recently gained much acclaim in the field of research as it has the capability to create high performance learners through virtual environments or by using data gathered from other domains. This systematic review defines (a) transfer learning; (b) discusses the recent research conducted; (c) the current status of transfer learning and finally, (d) discusses how transfer learning can bridge the gap between the virtual and the real.

Sign in / Sign up

Export Citation Format

Share Document