Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “deep”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,585 records · Page 88

GTC/OSIRIS SPECTROSCOPIC IDENTIFICATION OF A FAINT L SUBDWARF IN THE UKIRT INFRARED DEEP SKY SURVEY

We present the discovery of an L subdwarf in 234 deg{sup 2} common to the UK InfraRed Telescope (UKIRT) Infrared Deep Sky Survey Large Area Survey Data Release 2 and the Sloan Digital Sky Survey Data Release 3. This is the fifth L subdwarf announced to date, the first one identified in the UKIRT Infrared Deep Sky Survey, and the faintest known. The blue optical and near-infrared colors of ULAS J135058.86+081506.8 and its overall spectra energy distribution are similar to the known mid-L subdwarfs. Low-resolution optical (700-1000 nm) spectroscopy with the Optical System for Imaging and low Resolution Integrated Spectroscopy spectrograph on the 10.4 m Gran Telescopio de Canarias reveals that ULAS J135058.86+081506.8 exhibits a strong K I pressure-broadened line at 770 nm and a red slope longward of 800 nm, features characteristics of L-type dwarfs. From direct comparison with the four known L subdwarfs, we estimate its spectral type to be sdL4-sdL6 and derive a distance in the interval 94-170 pc. We provide a rough estimate of the space density for mid-L subdwarfs of 1.5 x 10{sup -4} pc{sup -3}.

79 ASTRONOMY AND ASTROPHYSICS↗

A SPECTROSCOPICALLY CONFIRMED X-RAY CLUSTER AT z = 1.62 WITH A POSSIBLE COMPANION IN THE SUBARU/XMM-NEWTON DEEP FIELD

We report on a confirmed galaxy cluster at z = 1.62. We discovered two concentrations of galaxies at z {approx} 1.6 in the Subaru/XMM-Newton deep field based on deep multi-band photometric data. We made a near-IR spectroscopic follow-up observation of them and confirmed several massive galaxies at z = 1.62. One of the two is associated with an extended X-ray emission at 4.5{sigma} on a scale of 0.'5, which is typical of high-z clusters. The X-ray detection suggests that it is a gravitationally bound system. The other one shows a hint of an X-ray signal, but only at 1.5{sigma}, and we obtained only one secure redshift at z = 1.62. We are not yet sure if this is a collapsed system. The possible twins exhibit a clear red sequence at K < 22 and seem to host relatively few number of faint red galaxies. Massive red galaxies are likely old galaxies-they have colors consistent with the formation redshift of z{sub f} = 3, and a spectral fit of the brightest confirmed member yields an age of 1.8{sup +0.1}{sub -0.2} Gyr with a mass of 2.5{sup +0.2}{sub -0.1} x 10{sup 11} M{sub sun}. Our results show that it is feasible to detect clusters at z > 1.5 in X-rays and also to perform detailed analysis of galaxies in them with the existing near-IR facilities on large telescopes.

79 ASTRONOMY AND ASTROPHYSICS↗

CONVECTIVE NATURE OF SUNSPOT PENUMBRAL FILAMENTS: DISCOVERY OF DOWNFLOWS IN THE DEEP PHOTOSPHERE

We study the velocity structure of penumbral filaments in the deep photosphere to obtain direct evidence for the convective nature of sunspot penumbrae. A sunspot was observed at high spatial resolution with the 1 m Swedish Solar Telescope in the deep photospheric C I 5380 A absorption line. The Multi-Object Multi-Frame Blind Deconvolution method is used for image restoration and straylight is filtered out. We report here the discovery of clear redshifts in the C I 5380 A line at multiple locations in sunspot penumbral filaments. For example, bright head of filaments show larger concentrated blueshift and are surrounded by darker, redshifted regions, suggestive of overturning convection. Elongated downflow lanes are also located beside bright penumbral fibrils. Our results provide the strongest evidence yet for the presence of overturning convection in penumbral filaments and highlight the need to observe the deepest layers of the penumbra in order to uncover the energy transport processes taking place there.

79 ASTRONOMY AND ASTROPHYSICS↗

MORPHOLOGICAL EVOLUTION OF GALAXIES FROM ULTRA-DEEP HUBBLE SPACE TELESCOPE WIDE FIELD CAMERA 3 IMAGING: THE HUBBLE SEQUENCE AT z {approx} 2

We use ultra-deep Hubble Space Telescope Wide Field Camera 3/infrared imaging of the Hubble Ultra Deep Field to investigate the rest-frame optical morphologies of a mass-selected sample of galaxies at z {approx} 2. We find a large variety of galaxy morphologies, ranging from large, blue, disk-like galaxies to compact, red, early-type galaxies. We derive rest-frame u - g color profiles for these galaxies and show that most z {approx} 2 galaxies in our sample have negative color gradients such that their cores are red. Although these color gradients may partly be caused by radial variations in dust content, they point to the existence of older stellar populations in the centers of z {approx} 2 galaxies. This result is consistent with an 'inside-out' scenario of galaxy growth. We find that the median color gradient is fairly constant with redshift: ({Delta}(u - g{sub rest})/{Delta}(log r)){sub median} = -0.47, -0.33, and -0.46 for z {approx} 2, z {approx} 1, and z = 0, respectively. Using structural parameters derived from surface brightness profiles we confirm that at z {approx} 2 galaxy morphology correlates well with specific star formation rate. At the same mass, star-forming galaxies have larger effective radii, bluer rest-frame u - g colors, and lower Sersic indices than quiescent galaxies. These correlations are very similar to those at lower redshift, suggesting that the relations that give rise to the Hubble sequence at z = 0 are already in place for massive galaxies at this early epoch.

79 ASTRONOMY AND ASTROPHYSICS↗

UNUSUAL LONG AND LUMINOUS OPTICAL TRANSIENT IN THE SUBARU DEEP FIELD

We present observations of SDF-05M05, an unusual optical transient discovered in the Subaru Deep Field (SDF). The duration of the transient is > {approx} 800 days in the observer frame, and the maximum brightness during observation reached approximately 23 mag in the i' and z' bands. The faint host galaxy is clearly identified in all five optical bands of the deep SDF images. The photometric redshift of the host yields z {approx} 0.6 and the corresponding absolute magnitude at maximum is {approx} - 20. This implies that this event shone with an absolute magnitude brighter than -19 mag for approximately 300 days in the rest frame, which is significantly longer than a typical supernova and ultraluminous supernova. The total radiated energy during our observation was 1 Multiplication-Sign 10{sup 51} erg. The light curves and color evolution are marginally consistent with some luminous IIn supernovae. We suggest that the transient may be a unique and peculiar supernova at intermediate redshift.

79 ASTRONOMY AND ASTROPHYSICS↗

THE ABUNDANCE OF STAR-FORMING GALAXIES IN THE REDSHIFT RANGE 8.5-12: NEW RESULTS FROM THE 2012 HUBBLE ULTRA DEEP FIELD CAMPAIGN

We present the results of the deepest search to date for star-forming galaxies beyond a redshift z {approx_equal} 8.5 utilizing a new sequence of near-infrared Wide-Field Camera 3 (WFC3/IR) images of the Hubble Ultra Deep Field (UDF). This 'UDF12' campaign completed in 2012 September doubles the earlier exposures with WFC3/IR in this field and quadruples the exposure in the key F105W filter used to locate such distant galaxies. Combined with additional imaging in the F140W filter, the fidelity of high-redshift candidates is greatly improved. Using spectral energy distribution fitting techniques on objects selected from a deep multi-band near-infrared stack, we find seven promising z > 8.5 candidates. As none of the previously claimed UDF candidates with 8.5 < z < 10 are confirmed by our deeper multi-band imaging, our campaign has transformed the measured abundance of galaxies in this redshift range. Although we recover the candidate UDFj-39546284 (previously proposed at z = 10.3), it is undetected in the newly added F140W image, implying that it lies at z = 11.9 or is an intense emission line galaxy at z {approx_equal} 2.4. Although no physically plausible model can explain the required line intensity given the lack of Ly{alpha} or broadband UV signal, without an infrared spectrum we cannot rule out an exotic interloper. Regardless, our robust z {approx_equal} 8.5-10 sample demonstrates a luminosity density that continues the smooth decline observed over 6 < z < 8. Such continuity has important implications for models of cosmic reionization and future searches for z >10 galaxies with James Webb Space Telescope.

79 ASTRONOMY AND ASTROPHYSICS↗

DEEP LBT/LUCI SPECTROSCOPY OF AN Ly{alpha} EMITTER CANDIDATE AT z {approx_equal} 7.7

We present deep spectroscopic observations of an Ly{alpha} emitter (LAE) candidate at z {approx_equal} 7.7 using the infrared spectrograph LUCI on the 2 Multiplication-Sign 8.4 m Large Binocular Telescope (LBT). The candidate is the brightest among the four z {approx_equal} 7.7 LAE candidates found in a narrowband imaging survey by Krug et al. Our spectroscopic data include a total of 7.5 hr of integration with LBT/LUCI and are deep enough to significantly (3.2{sigma}-4.9{sigma}) detect the Ly{alpha} emission line of this candidate based on its Ly{alpha} flux 1.2 Multiplication-Sign 10{sup -17} erg s{sup -1} cm{sup -2} estimated from the narrowband photometry. However, we do not find any convincing signal at the expected position of its Ly{alpha} emission line, suggesting that this source is not an LAE at z {approx_equal} 7.7. The non-detection in this work, together with the previous studies of z {approx_equal} 7.7 LAEs, puts a strong constraint on the bright-end Ly{alpha} luminosity function (LF) at z {approx_equal} 7.7. We find a rapid evolution of the Ly{alpha} LF from z {approx_equal} 6.5 to 7.7: the upper limit of the z {approx_equal} 7.7 LF is more than five times lower than the z {approx_equal} 6.5 LF at the bright end (f{>=} 1.0 Multiplication-Sign 10{sup -17} erg s{sup -1} cm{sup -2} or L{>=} 6.9 Multiplication-Sign 10{sup 42} erg s{sup -1}). This is likely caused by an increasing neutral fraction in the intergalactic medium that substantially attenuates Ly{alpha} emission at z {approx_equal} 7.7.

79 ASTRONOMY AND ASTROPHYSICS↗

THE SPECTRAL ENERGY DISTRIBUTIONS OF z ∼ 8 GALAXIES FROM THE IRAC ULTRA DEEP FIELDS: EMISSION LINES, STELLAR MASSES, AND SPECIFIC STAR FORMATION RATES AT 650 MYR

Using new ultradeep Spitzer/InfraRed Array Camera (IRAC) photometry from the IRAC Ultra Deep Field program, we investigate the stellar populations of a sample of 63 Y-dropout galaxy candidates at z ∼ 8, only 650 Myr after the big bang. The sources are selected from HST/ACS+WFC3/IR data over the Hubble Ultra Deep Field (HUDF), two HUDF parallel fields, and wide area data over the CANDELS/GOODS-South. The new Spitzer/IRAC data increase the coverage in [3.6] and [4.5] to ∼120h over the HUDF reaching depths of ∼28 (AB,1σ). The improved depth and inclusion of brighter candidates result in direct ≥3σ InfraRed Array Camera (IRAC) detections of 20/63 sources, of which 11/63 are detected at ≥5σ. The average [3.6]-[4.5] colors of IRAC detected galaxies at z ∼ 8 are markedly redder than those at z ∼ 7, observed only 130 Myr later. The simplest explanation is that we witness strong rest-frame optical emission lines (in particular [O III] λλ4959, 5007 + Hβ) moving through the IRAC bandpasses with redshift. Assuming that the average rest-frame spectrum is the same at both z ∼ 7 and z ∼ 8 we estimate a rest-frame equivalent width of contributing 0.56{sup +0.16}{sub -0.11} mag to the [4.5] filter at z ∼ 8. The corresponding W{sub Hα}=430{sup +160}{sub -110} Å implies an average specific star formation rate of sSFR=11{sub -5}{sup +11} Gyr{sup –1} and a stellar population age of 100{sub -50}{sup +100} Myr. Correcting the spectral energy distribution for the contribution of emission lines lowers the average best-fit stellar masses and mass-to-light ratios by ∼3 ×, decreasing the integrated stellar mass density to ρ{sup *}(z=8,M{sub UV}<-18)=0.6{sup +0.4}{sub -0.3}×10{sup 6} M{sub sun} Mpc{sup –3}.

79 ASTRONOMY AND ASTROPHYSICS↗

The deep-acceptor nature of the chalcogen vacancies in 2D transition-metal dichalcogenides

Abstract Chalcogen vacancies in the semiconducting monolayer transition-metal dichalcogenides (TMDs) have frequently been invoked to explain a wide range of phenomena, including both unintentional p-type and n-type conductivity, as well as sub-band gap defect levels measured via tunneling or optical spectroscopy. These conflicting interpretations of the deep versus shallow nature of the chalcogen vacancies are due in part to shortcomings in prior first-principles calculations of defects in the semiconducting two-dimensional TMDs that have been used to explain experimental observations. Here we report results of hybrid density functional calculations for the chalcogen vacancy in a series of monolayer TMDs, correctly referencing the thermodynamic charge transition levels to the fundamental band gap (as opposed to the optical band gap). We find that the chalcogen vacancies are deep acceptors and cannot lead to n-type or p-type conductivity. Both the (0/−1) and (−1/−2) transition levels occur in the gap, leading to paramagnetic charge states S = 1 / 2 and S = 1, respectively, in a collinear-spin representation. We discuss trends in terms of the band alignments between the TMDs, which can serve as a guide to future experimental studies of vacancy behavior.

36 MATERIALS SCIENCE↗

Deeply uncertain: comparing methods of uncertainty quantification in deep learning algorithms

We present a comparison of methods for uncertainty quantification (UQ) in deep learning algorithms in the context of a simple physical system. Three of the most common uncertainty quantification methods - Bayesian Neural Networks (BNN), Concrete Dropout (CD), and Deep Ensembles (DE) - are compared to the standard analytic error propagation. We discuss this comparison in terms endemic to both machine learning ("epistemic" and "aleatoric") and the physical sciences ("statistical" and "systematic"). The comparisons are presented in terms of simulated experimental measurements of a single pendulum - a prototypical physical system for studying measurement and analysis techniques. Our results highlight some pitfalls that may occur when using these UQ methods. For example, when the variation of noise in the training set is small, all methods predicted the same relative uncertainty independently of the inputs. This issue is particularly hard to avoid in BNN. On the other hand, when the test set contains samples far from the training distribution, we found that no methods sufficiently increased the uncertainties associated to their predictions. This problem was particularly clear for CD. In light of these results, we make some recommendations for usage and interpretation of UQ methods.

59 BASIC BIOLOGICAL SCIENCES↗

Distributed deep reinforcement learning for simulation control

Abstract Several applications in the scientific simulation of physical systems can be formulated as control/optimization problems. The computational models for such systems generally contain hyperparameters, which control solution fidelity and computational expense. The tuning of these parameters is non-trivial and the general approach is to manually ‘spot-check’ for good combinations. This is because optimal hyperparameter configuration search becomes intractable when the parameter space is large and when they may vary dynamically. To address this issue, we present a framework based on deep reinforcement learning (RL) to train a deep neural network agent that controls a model solve by varying parameters dynamically. First, we validate our RL framework for the problem of controlling chaos in chaotic systems by dynamically changing the parameters of the system. Subsequently, we illustrate the capabilities of our framework for accelerating the convergence of a steady-state computational fluid dynamics solver by automatically adjusting the relaxation factors of the discretized Navier–Stokes equations during run-time. The results indicate that the run-time control of the relaxation factors by the learned policy leads to a significant reduction in the number of iterations for convergence compared to the random selection of the relaxation factors. Our results point to potential benefits for learning adaptive hyperparameter learning strategies across different geometries and boundary conditions with implications for reduced computational campaign expenses 4 4 Data and codes available at https://github.com/Romit-Maulik/PAR-RL . .

42 ENGINEERING↗

Evaluating probabilistic deep learning methods for uncertainty quantification of temperature downscaling

Deep learning (DL) has emerged as a promising tool for downscaling coarse-resolution climate data to high-resolution outputs, enabling improved regional climate predictions. A critical aspect of DL-based downscaling is the incorporation of uncertainty quantification (UQ), which enhances the interpretability and reliability of predictions—key factors for climate risk assessment and decision-making. This study develops a DL model to downscale 2 m temperature across the contiguous United States using reanalysis datasets. We systematically evaluate three epistemic UQ methods—deep ensembles (DEns), Monte Carlo dropout (MCD), and Flipout—based on their probabilistic accuracy, downscaling performance, sensitivity to geographical features, and computational efficiency. Results indicate that MCD generally outperforms Flipout and DEns in terms of calibration and downscaling accuracy. However, DEns demonstrate lower calibration errors in coastal regions, indicating its higher confidence within these areas. Flipout, in contrast, is more sensitive to elevation gradients and exhibits higher calibration errors in mountainous regions. Hence, the choice of UQ method for this task depends on the specific requirements of the application. For applications that prioritize overall calibration, downscaling accuracy, and computational efficiency, MCD is a strong candidate. These findings highlight the importance of selecting UQ methods based on application-specific requirements, such as geographical context and computational constraints. By addressing the trade-offs between UQ methods, this study provides actionable insights for improving the reliability, scalability, and utility of DL-based downscaling in climate science.

Environmental sciences↗

Data imbalance in drug response prediction: multi-objective optimization approach in deep learning setting

Abstract Drug response prediction (DRP) methods tackle the complex task of associating the effectiveness of small molecules with the specific genetic makeup of the patient. Anti-cancer DRP is a particularly challenging task requiring costly experiments as underlying pathogenic mechanisms are broad and associated with multiple genomic pathways. The scientific community has exerted significant efforts to generate public drug screening datasets, giving a path to various machine learning models that attempt to reason over complex data space of small compounds and biological characteristics of tumors. However, the data depth is still lacking compared to application domains like computer vision or natural language processing domains, limiting current learning capabilities. To combat this issue and improves the generalizability of the DRP models, we are exploring strategies that explicitly address the imbalance in the DRP datasets. We reframe the problem as a multi-objective optimization across multiple drugs to maximize deep learning model performance. We implement this approach by constructing Multi-Objective Optimization Regularized by Loss Entropy loss function and plugging it into a Deep Learning model. We demonstrate the utility of proposed drug discovery methods and make suggestions for further potential application of the work to achieve desirable outcomes in the healthcare field.

Biochemistry & Molecular Biology↗

A Spitzer survey of Deep Drilling Fields to be targeted by the Vera C. Rubin Observatory Legacy Survey of Space and Time

The Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST) will observe several Deep Drilling Fields (DDFs) to a greater depth and with a more rapid cadence than the main survey. In this paper, we describe the ‘DeepDrill’ survey, which used the Spitzer Space Telescope Infrared Array Camera (IRAC) to observe three of the four currently defined DDFs in two bands, centred on 3.6 and 4.5 μm. Furthermore, these observations expand the area that was covered by an earlier set of observations in these three fields by the Spitzer Extragalactic Representative Volume Survey (SERVS). The combined DeepDrill and SERVS data cover the footprints of the LSST DDFs in the Extended Chandra Deep Field–South (ECDFS) field, the ELAIS-S1 field (ES1), and the XMM-Large-Scale Structure Survey field (XMM-LSS). The observations reach an approximate 5σ point-source depth of 2 μJy (corresponding to an AB magnitude of 23.1; sufficient to detect a 10$^{11} \, \mathrm{M}_{\odot}$ galaxy out to z ≈ 5) in each of the two bands over a total area of $\approx 29\,$ deg2. The dual-band catalogues contain a total of 2.35 million sources. In this paper, we describe the observations and data products from the survey, and an overview of the properties of galaxies in the survey. We compare the source counts to predictions from the Shark semi-analytic model of galaxy formation. We also identify a population of sources with extremely red ([3.6]-[4.5] >1.2) colours which we show mostly consists of highly obscured active galactic nuclei.

79 ASTRONOMY AND ASTROPHYSICS↗

Peculiar velocity estimation from kinetic SZ effect using deep neural networks

ABSTRACT The Sunyaev–Zel’dolvich (SZ) effect is expected to be instrumental in measuring velocities of distant clusters in near future telescope surveys. We simplify the calculation of peculiar velocities of galaxy clusters using deep learning frameworks trained on numerical simulations to avoid the independent estimation of the optical depth. Images of distorted photon backgrounds are generated for idealized observations using one of the largest cosmological hydrodynamical simulations, the Magneticum simulations. The model is tested to determine its ability of estimating peculiar velocities from future kinetic SZ observations under different noise conditions. The deep learning algorithm displays robustness in estimating peculiar velocities from kinetic SZ effect by an improvement in accuracy of about 17 per cent compared to the analytical approach.

Wang, Yuyu (ORCID:0000000202458547)↗

Ionizing radiation from AGNs at z > 3.3 with the Subaru Hyper Suprime-Cam Survey and the CFHT Large Area U-band Deep Survey (CLAUDS)

ABSTRACT We use deep and wide imaging data from the CFHT Large Area U-band Deep Survey (CLAUDS) and the Hyper Suprime-Cam Subaru Strategic Program (HSC-SSP) to constrain the ionizing radiation (Lyman Continuum, LyC) escape fraction from active galactic nuclei (AGNs) at z ∼ 3–4. For 94 AGNs with spectroscopic redshifts at 3.3 < z < 4.0, we use their U/i-band flux ratios to estimate LyC transmission of individual AGNs. The distribution of their LyC transmission shows values lower than the range of LyC transmission values for IGM of the same redshift range, which suggests that LyC escape fraction of AGNs at z > 3.3 is considerably lower than unity in most cases. We do not find any trend in LyC transmission values depending on their UV luminosities. Based on the photometry of stacked images we find the average flux ratio of LyC and non-ionizing UV photons escaping from the objects (fLyC/fUV)out = 0.182 ± 0.043 for AGNs at 3.3 < z < 3.6, which corresponds to LyC escape fraction fesc = 0.303 ± 0.072 if we assume a fiducial intrinsic SED of AGN. Based on the estimated LyC escape fraction and the UV luminosity function (UVLF) of AGNs, we argue that UV-selected AGNs’ contribution to the LyC emissivity at the epoch is minor, although the size of their contribution largely depends on the shape of the UVLF.

79 ASTRONOMY AND ASTROPHYSICS↗

SDSS-IV DR17: final release of MaNGA PyMorph photometric and deep-learning morphological catalogues

ABSTRACT We present the MaNGA PyMorph photometric Value Added Catalogue (MPP-VAC-DR17) and the MaNGA Deep Learning Morphological VAC (MDLM-VAC-DR17) for the final data release of the MaNGA survey, which is part of the SDSS Data Release 17 (DR17). The MPP-VAC-DR17 provides photometric parameters from Sérsic and Sérsic+Exponential fits to the two-dimensional surface brightness profiles of the MaNGA DR17 galaxy sample in the g, r, and i bands (e.g. total fluxes, half-light radii, bulge-disc fractions, ellipticities, position angles, etc.). The MDLM-VAC-DR17 provides deep-learning-based morphological classifications for the same galaxies. The MDLM-VAC-DR17 includes a number of morphological properties, for example, a T-Type, a finer separation between elliptical and S0, as well as the identification of edge-on and barred galaxies. While the MPP-VAC-DR17 simply extends the MaNGA PyMorph photometric VAC published in the SDSS Data Release 15 (MPP-VAC-DR15) to now include galaxies that were added to make the final DR17, the MDLM-VAC-DR17 implements some changes and improvements compared to the previous release (MDLM-VAC-DR15): Namely, the low end of the T-Types is better recovered in this new version. The catalogue also includes a separation between early or late type, which classifies the two populations in a complementary way to the T-Type, especially at the intermediate types (−1 < T-Type < 2), where the T-Type values show a large scatter. In addition, k-fold-based uncertainties on the classifications are also provided. To ensure robustness and reliability, we have also visually inspected all the images. We describe the content of the catalogues and show some interesting ways in which they can be combined.

79 ASTRONOMY AND ASTROPHYSICS↗

Predicting images for the dynamics of stellar clusters ( π-DOC ): a deep learning framework to predict mass, distance, and age of globular clusters

ABSTRACT Dynamical mass estimates of simple systems such as globular clusters (GCs) still suffer from up to a factor of 2 uncertainty. This is primarily due to the oversimplifications of standard dynamical models that often neglect the effects of the long-term evolution of GCs. Here, we introduce a new approach to measure the dynamical properties of GCs, based on the combination of a deep-learning framework and the large amount of data from direct N-body simulations. Our algorithm, π-DOC (Predicting Images for the Dynamics Of stellar Clusters) is composed of two convolutional networks, trained to learn the non-trivial transformation between an observed GC luminosity map and its associated mass distribution, age, and distance. The training set is made of V-band luminosity and mass maps constructed as mock observations from N-body simulations. The tests on π-DOC demonstrate that we can predict the mass distribution with a mean error per pixel of 27 per cent, and the age and distance with an accuracy of 1.5 Gyr and 6 kpc, respectively. In turn, we recover the shape of the mass-to-light profile and its global value with a mean error of 12 per cent, which implies that we efficiently trace mass segregation. A preliminary comparison with observations indicates that our algorithm is able to predict the dynamical properties of GCs within the limits of the training set. These encouraging results demonstrate that our deep-learning framework and its forward modelling approach can offer a rapid and adaptable tool competitive with standard dynamical models.

Chardin, Jonathan↗