ABSTRACT
In this work, we explore the possibility of applying machine learning methods designed for 1D problems to the task of galaxy image classification. The algorithms used for image ...classification typically rely on multiple costly steps, such as the point spread function deconvolution and the training and application of complex Convolutional Neural Networks of thousands or even millions of parameters. In our approach, we extract features from the galaxy images by analysing the elliptical isophotes in their light distribution and collect the information in a sequence. The sequences obtained with this method present definite features allowing a direct distinction between galaxy types. Then, we train and classify the sequences with machine learning algorithms, designed through the platform Modulos AutoML. As a demonstration of this method, we use the second public release of the Dark Energy Survey (DES DR2). We show that we are able to successfully distinguish between early-type and late-type galaxies, for images with signal-to-noise ratio greater than 300. This yields an accuracy of $86{{\ \rm per\ cent}}$ for the early-type galaxies and $93{{\ \rm per\ cent}}$ for the late-type galaxies, which is on par with most contemporary automated image classification approaches. The data dimensionality reduction of our novel method implies a significant lowering in computational cost of classification. In the perspective of future data sets obtained with e.g. Euclid and the Vera Rubin Observatory, this work represents a path towards using a well-tested and widely used platform from industry in efficiently tackling galaxy classification problems at the peta-byte scale.
We present a forward-modeling simulation framework designed to model the data products from the Dark Energy Survey (DES). This forward-model process can be thought of as a transfer function-a mapping ...from cosmological/astronomical signals to the final data products used by the scientists. Using output from the cosmological simulations (the Blind Cosmology Challenge), we generate simulated images (the Ultra Fast Image Simulator) and catalogs representative of the DES data. In this work we demonstrate the framework by simulating the 244 deg super(2) coadd images and catalogs in five bands for the DES Science Verification data. The simulation output is compared with the corresponding data to show that major characteristics of the images and catalogs can be captured. We also point out several directions of future improvements. Two practical examples-star-galaxy classification and proximity effects on object detection-are then used to illustrate how one can use the simulations to address systematics issues in data analysis. With clear understanding of the simplifications in our model, we show that one can use the simulations side-by-side with data products to interpret the measurements. This forward modeling approach is generally applicable for other upcoming and future surveys. It provides a powerful tool for systematics studies that is sufficiently realistic and highly controllable.
The purpose of this work is to model dispersion at the macroscale as the result of mechanisms acting at the microscale. We simulated advection and molecular diffusion of a passive tracer in networks ...of capillaries with different radius distributions. The solute transport process asymptotically reached a Fickian regime at the end of a transitory period of variable duration. One effect of heterogeneity was to increase the asymptotic dispersion coefficients and the transient time. Another, more important, result is the observation of a transition of the longitudinal dispersion from the Taylor‐Aris dispersion in homogeneous networks to the so‐called mechanical dispersion in highly heterogeneous ones. The existence of this transition is supported by previous numerical and experimental studies. In the case of transverse dispersion, the asymptotic transverse coefficient appeared very weakly related to the Peclet number at all heterogeneity levels. We propose an explanation to this apparent contradiction to experimental observations.
ABSTRACT In this work, we explore the possibility of applying machine learning methods designed for 1D problems to the task of galaxy image classification. The algorithms used for image ...classification typically rely on multiple costly steps, such as the point spread function deconvolution and the training and application of complex Convolutional Neural Networks of thousands or even millions of parameters. In our approach, we extract features from the galaxy images by analysing the elliptical isophotes in their light distribution and collect the information in a sequence. The sequences obtained with this method present definite features allowing a direct distinction between galaxy types. Then, we train and classify the sequences with machine learning algorithms, designed through the platform Modulos AutoML. As a demonstration of this method, we use the second public release of the Dark Energy Survey (DES DR2). We show that we are able to successfully distinguish between early-type and late-type galaxies, for images with signal-to-noise ratio greater than 300. This yields an accuracy of $86{{\ \rm per\ cent}}$ for the early-type galaxies and $93{{\ \rm per\ cent}}$ for the late-type galaxies, which is on par with most contemporary automated image classification approaches. The data dimensionality reduction of our novel method implies a significant lowering in computational cost of classification. In the perspective of future data sets obtained with e.g. Euclid and the Vera Rubin Observatory, this work represents a path towards using a well-tested and widely used platform from industry in efficiently tackling galaxy classification problems at the peta-byte scale.
We present photometric redshift estimates for galaxies used in the weak lensing analysis of the Dark Energy Survey Science Verification (DES SV) data. Four model- or machine learning-based ...photometric redshift methods-annz2, bpz calibrated against BCC-Ufig simulations, skynet, and tpz-are analyzed. For training, calibration, and testing of these methods, we construct a catalogue of spectroscopically confirmed galaxies matched against DES SV data. The performance of the methods is evaluated against the matched spectroscopic catalogue, focusing on metrics relevant for weak lensing analyses, with additional validation against COSMOS photo-z's. From the galaxies in the DES SV shear catalogue, which have mean redshift 0.72 + or - 0.01 over the range 0.3 < z< 1.3, we construct three tomographic bins with means of z= {0.45,0.67,1.00}. These bins each have systematic uncertainties delta sub(z)<, ~ 0.05 in the mean of the fiducial skynet photo-z n(z). We propagate the errors in the redshift distributions through to their impact on cosmological parameters estimated with cosmic shear, and find that they cause shifts in the value of sigma sub(8) of approximately 3%. This shift is within the one sigma statistical errors on sigma sub(8) for the DES SV shear catalogue. We further study the potential impact of systematic differences on the critical surface density, capital sigma sub(crit), finding levels of bias safely less than the statistical power of DES SV data. We recommend a final Gaussian prior for the photo-z bias in the mean of n(z) of width 0.05 for each of the three tomographic bins, and show that this is a sufficient bias model for the corresponding cosmology analysis.
ABSTRACT Spatially varying depth and the characteristics of observing conditions, such as seeing, airmass, or sky background, are major sources of systematic uncertainties in modern galaxy survey ...analyses, particularly in deep multi-epoch surveys. We present a framework to extract and project these sources of systematics onto the sky, and apply it to the Dark Energy Survey (DES) to map the observing conditions of the Science Verification (SV) data. The resulting distributions and maps of sources of systematics are used in several analyses of DES-SV to perform detailed null tests with the data, and also to incorporate systematics in survey simulations. We illustrate the complementary nature of these two approaches by comparing the SV data with BCC-UFig, a synthetic sky catalog generated by forward-modeling of the DES-SV images. We analyze the BCC-UFig simulation to construct galaxy samples mimicking those used in SV galaxy clustering studies. We show that the spatially varying survey depth imprinted in the observed galaxy densities and the redshift distributions of the SV data are successfully reproduced by the simulation and are well-captured by the maps of observing conditions. The combined use of the maps, the SV data, and the BCC-UFig simulation allows us to quantify the impact of spatial systematics on N(z), the redshift distributions inferred using photometric redshifts. We conclude that spatial systematics in the SV data are mainly due to seeing fluctuations and are under control in current clustering and weak-lensing analyses. However, they will need to be carefully characterized in upcoming phases of DES in order to avoid biasing the inferred cosmological results. The framework presented here is relevant to all multi-epoch surveys and will be essential for exploiting future surveys such as the Large Synoptic Survey Telescope, which will require detailed null tests and realistic end-to-end image simulations to correctly interpret the deep, high-cadence observations of the sky.
We present a structural and morphological catalogue for 45 million objects selected from the first year data of the Dark Energy Survey (DES). Single Sérsic fits and non-parametric measurements are ...produced for g, r, and i filters. The parameters from the best-fitting Sérsic model (total magnitude, half-light radius, Sérsic index, axis ratio, and position angle) are measured with galfit; the non-parametric coefficients (concentration, asymmetry, clumpiness, Gini, M20) are provided using the Zurich Estimator of Structural Types (zest+). To study the statistical uncertainties, we consider a sample of state-of-the-art image simulations with a realistic distribution in the input parameter space and then process and analyse them as we do with real data: this enables us to quantify the observational biases due to PSF blurring and magnitude effects and correct the measurements as a function of magnitude, galaxy size, Sérsic index (concentration for the analysis of the non-parametric measurements) and ellipticity. We present the largest structural catalogue to date: we find that accurate and complete measurements for all the structural parameters are typically obtained for galaxies with SExtractorMAG_AUTO_I ≤ 21. Indeed, the parameters in the filters i and r can be overall well recovered up to MAG_AUTO ≤ 21.5, corresponding to a fitting completeness of ~90 per cent below this threshold, for a total of 25 million galaxies. The combination of parametric and non-parametric structural measurements makes this catalogue an important instrument to explore and understand how galaxies form and evolve. The catalogue described in this paper will be publicly released alongside the DES collaboration Y1 cosmology data products at the following URL: https://des.ncsa.illinois.edu/releases.
Cosmic shear is sensitive to fluctuations in the cosmological matter density field, including on small physical scales, where matter clustering is affected by baryonic physics in galaxies and galaxy ...clusters, such as star formation, supernovae feedback, and active galactic nuclei feedback. While muddying any cosmological information that is contained in small-scale cosmic shear measurements, this does mean that cosmic shear has the potential to constrain baryonic physics and galaxy formation. We perform an analysis of the Dark Energy Survey (DES) Science Verification (SV) cosmic shear measurements, now extended to smaller scales, and using the Mead et al. (2015) halo model to account for baryonic feedback. While the SV data has limited statistical power, we demonstrate using a simulated likelihood analysis that the final DES data will have the statistical power to differentiate among baryonic feedback scenarios. We also explore some of the difficulties in interpreting the small scales in cosmic shear measurements, presenting estimates of the size of several other systematic effects that make inference from small scales difficult, including uncertainty in the modelling of intrinsic alignment on non-linear scales, 'lensing bias', and shape measurement selection effects. For the latter two, we make use of novel image simulations. While future cosmic shear data sets have the statistical power to constrain baryonic feedback scenarios, there are several systematic effects that require improved treatments, in order to make robust conclusions about baryonic feedback.
The changes of fracture surfaces geometry and extend are studied using X‐ray tomography during aperture increase due to CO2‐rich fluid percolation. Dissolution experiments were conducted on two ...micritic rock samples; one pure calcite end‐member and one with typical composition for marine carbonates (85% calcite). High‐resolution digital images of the fracture geometry allow quantifying the surface properties changes over four spatial scales with a resolution of 4.91 μm. Fracture surfaces are self‐affine with an initial dimension of 2.5. Dissolution of the pure‐calcite sample is clearly a process of homogeneous chemical “erosion” of the surface elevation: fractal dimension and specific surface remains constant (1.5 times the planar surface). Conversely, for the 85% calcite sample, initial topographic surfaces of the fracture walls evolve rapidly toward “non‐topographic” interfaces displaying overhangs due the preferential dissolution of the carbonate grains. In this case, the conventional definition of the effective aperture must be revisited. Such structures can only be assessed from 3D observations. As dissolution progresses, the specific surface increases strongly, more than 5 times the planar surface, and probably faster than the reactive surface.
The goal of this paper is to evaluate the effect of the variance of pore size distribution on the transport properties of rocks. Several heterogeneous network realizations with very broad, uniform, ...or log uniform pore size distributions were constructed. A series of networks were then derived deterministically from these initial networks by repeatedly applying a shrinking operator to the pores of the original realizations. This operator was devised in such a way as to maintain the mean pore size constant while changing the variance of the pore size distribution, therefore allowing its effect on the transport properties to be isolated. We thus assessed the validity of several permeability models from the literature as a function of the variance of the pore size distribution. We found that the product of the permeability by the electrical formation factor was proportional to the square of the critical radius as proposed in the Katz‐Thompson model Katz and Thompson, 1986. However, we observed that the dramatic flow localization occurring at high pore size variance was not restricted to the backbone of the critical subnetwork (or critical path) as assumed in the Katz‐Thompson model. We propose that a better justification of the relation mentioned above arises from the underlying percolation problem of the viscous‐inertial transition observed in harmonic flow as a function of frequency. In addition, we appraised the stochastic Bernabé‐Revil model Bernabé and Revil, 1995. We found that this model was more and more difficult to implement as the pore size variance was increased. A possible interpretation could be that at high levels of pore‐scale heterogeneity, very large pore size fluctuations occur and the flow pattern is so strongly and determimstically related to these extreme fluctuations that a stochastic description becomes inadequate.