arXiv subjects

Hossen Teimoorinia

Publications and source records attributed to Hossen Teimoorinia.

At least 19 recordsLinked to original sources

arXiv2024

Revisiting AGN Placement on the BPT Diagram: A Spectral Decomposition Approach

Traditional single-fibre spectroscopy provides a single galaxy spectrum, forming the basis for crucial parameter estimation. However, its accuracy can be compromised by various sources of contamination, such as the prominent \Ha~emission line originating from both Star-Forming (SF) regions and non-Star-Forming regions (NonSF), including Active Galactic Nuclei (AGN). The potential to dissect a spectrum into its SF and NonSF constituents holds the promise of significantly enhancing precision in parameter estimates. In contrast, Integral Field Unit (IFU) surveys present a solution to minimize contamination. These surveys examine spatially localized regions within galaxies, reducing the impact of mixed sources. Although an IFU survey's resulting spectrum covers a smaller region of a galaxy than single-fibre spectroscopy, it can still encompass a blend of heterogeneous sources. Our study introduces an innovative model informed by insights from the MaNGA IFU survey. This model enables the decomposition of galaxy spectra, including those from the Sloan Digital Sky Survey (SDSS), into SF and NonSF components. Applying our model to these survey datasets produces two distinct spectra, one for SF and another for NonSF components, while conserving flux across wavelength bins. When these decomposed spectra are visualized on a BPT diagram, interesting patterns emerge. There is a significant shift in the placement of the NonSF decomposed spectra, as well as the emergence of two distinct clusters in the LINER and Seyfert regions. This shift highlights the key role of SF `contamination' in influencing the positioning of NonSF spectra within the BPT diagram.

astro-ph.GA↗

arXiv2024

Stellar Populations With Optical Spectra: Deep Learning vs. Popular Spectrum Fitting Codes

We compare the performance of several popular spectrum fitting codes (Firefly, starlight, pyPipe3D and pPXF), and a deep-learning convolutional neural network (StarNet), in recovering known stellar population properties (mean stellar age, stellar metallicity, stellar mass-to-light ratio M*/L_r and the internal E(B-V)) of simulated galaxy spectra in optical wavelengths. Our mock spectra are constructed from star-formation histories from the IllustrisTNG100-1 simulation. These spectra mimic the Sloan Digital Sky Survey (SDSS) through a novel method of including the noise, sky residuals and emission lines taken directly from SDSS. We find that StarNet vastly outperforms all conventional codes in both speed and recovery of stellar population properties (error scatter < 0.08 dex, average biases < 0.02 dex for all tested quantities), but it requires an appropriate training set. Of the non-machine-learning codes, pPXF was a factor of 3-4 times faster than the other codes, and was the best in recovering stellar population properties (error scatter of < 0.11 dex, average biases < 0.08 dex). However, the errors and biases are strongly dependent on both true and predicted values of stellar age and metallicity, and signal-to-noise ratio. The biases of all codes can approach 0.15 dex in stellar ages, metallicities and log M*/L_r , but remain < 0.05 for E(B-V). Using unrealistic Gaussian noise in the construction of mock spectra will underestimate the errors in the metallicities by a factor of two or more, and mocks without emission lines will underestimate the errors in stellar age and M*/L_r by a factor of two.

astro-ph.GA↗

arXiv2023

Mitigating the Non-Linearities in a Pyramid Wavefront Sensor

For natural guide start adaptive optics (AO) systems, pyramid wavefront sensors (PWFSs) can provide significant increase in sensitivity over the traditional Shack-Hartmann, but at the cost of a reduced linear range. When using a linear reconstructor, non-linearities result in wavefront estimation errors, which can have a significant impact on the image quality delivered by the AO system. Here we simulate a wavefront passing through a PWFS under varying observing conditions to explore the possibility of using a non-linear machine learning model to estimate wavefront errors better than a linear reconstruction. We find significant improvement even with light-weight models, underscoring the need for further investigation of this approach.

astro-ph.IM↗

arXiv2022

Forecasting Wavefront Corrections in an Adaptive Optics System

We use telemetry data from the Gemini North ALTAIR adaptive optics system to investigate how well the commands for wavefront correction (both Tip/Tilt and high-order turbulence) can be forecasted in order to reduce lag error (due to wavefront sensor averaging and computational delays) and improve delivered image quality. We show that a high level of reduction ($\sim$ 5 for Tip-Tilt and $\sim$ 2 for high-order modes) in RMS wavefront error can be achieved by using a "forecasting filter" based on a linear auto-regressive model with only a few coefficients ($\sim$ 30 for Tip-Tilt and $\sim$ 5 for high-order modes) to complement the existing integral servo-controller. Updating this filter to adapt to evolving observing conditions is computationally inexpensive and requires less than 10 seconds worth of telemetry data. We also use several machine learning models (Long-Short Term Memory and dilated convolutional models) to evaluate whether further improvements could be achieved with a more sophisticated non-linear model. Our attempts showed no perceptible improvements over linear auto-regressive predictions, even for large lags where residuals from the linear models are high, suggesting that non-linear wavefront distortions for ALTAIR at the Gemini North telescope may not be forecasted with the current setup

astro-ph.IM↗

arXiv2022

The combined and respective roles of imaging and stellar kinematics in identifying galaxy merger remnants

One of the central challenges to establishing the role of mergers in galaxy evolution is the selection of pure and complete merger samples in observations. In particular, while large and reasonably pure interacting galaxy pair samples can be obtained with relative ease via spectroscopic criteria, automated selection of post-coalescence merger remnants is restricted to the physical characteristics of remnants alone. Furthermore, such selection has predominantly focused on imaging data -- whereas kinematic data may offer a complimentary basis for identifying merger remnants. Therefore, we examine the theoretical utility of both the morphological and kinematic features of merger remnants in distinguishing galaxy merger remnants from other galaxies. Deep classification models are calibrated and evaluated using idealized synthetic images and line-of-sight stellar velocity maps of a heterogeneous population of galaxies and merger remnants from the TNG100 cosmological hydrodynamical simulation. We show that even idealized stellar kinematic data has limited utility compared to imaging and under-performs by $2.1\%\pm0.5\%$ in completeness and $4.7\%\pm0.4\%$ in purity for our fiducial model architecture. Combining imaging and stellar kinematics offers a small boost in completeness (by $1.8\%\pm0.4\%$, compared to $92.7\%\pm0.2\%$ from imaging alone) but no change in purity ($0.1\%\pm0.3\%$ improvement compared to $92.7\%\pm0.2\%$, evaluated with equal numbers of merger remnant and non-remnant control galaxies). Classification accuracy of all models is particularly sensitive to physical companions at separations $\lesssim40$ kpc and to time-since-coalescence. Taken together, our results show that the stellar kinematic data has little to offer in compliment to imaging for merger remnant identification in a heterogeneous galaxy population.

astro-ph.GA↗

arXiv2021

Mapping the Diversity of Galaxy Spectra with Deep Unsupervised Machine Learning

Modern spectroscopic surveys of galaxies such as MaNGA consist of millions of diverse spectra covering different regions of thousands of galaxies. We propose and implement a deep unsupervised machine learning method to summarize the entire diversity of MaNGA spectra onto a 15x15 map (DESOM-1), where neighbouring points on the map represent similar spectra. We demonstrate our method as an alternative to conventional full spectral fitting for deriving physical quantities, as well as their full probability distributions, much more efficiently than traditional resource-intensive Bayesian methods. Since spectra are grouped by similarity, the distribution of spectra onto the map for a single galaxy, i.e., its "fingerprint", reveals the presence of distinct stellar populations within the galaxy indicating smoother or episodic star-formation histories. We further map the diversity of galaxy fingerprints onto a second map (DESOM-2). Using galaxy images and independent measures of galaxy morphology, we confirm that galaxies with similar fingerprints have similar morphologies and inclination angles. Since morphological information was not used in the mapping algorithm, relating galaxy morphology to the star-formation histories encoded in the fingerprints is one example of how the DESOM maps can be used to make scientific inferences.

astro-ph.GA↗

arXiv2021

Convolutional neural network identification of galaxy post-mergers in UNIONS using IllustrisTNG

The Canada-France Imaging Survey (CFIS) will consist of deep, high-resolution r-band imaging over ~5000 square degrees of the sky, representing a first-rate opportunity to identify recently-merged galaxies. Due to the large number of galaxies in CFIS, we investigate the use of a convolutional neural network (CNN) for automated merger classification. Training samples of post-merger and isolated galaxy images are generated from the IllustrisTNG simulation processed with the observational realism code RealSim. The CNN's overall classification accuracy is 88 percent, remaining stable over a wide range of intrinsic and environmental parameters. We generate a mock galaxy survey from IllustrisTNG in order to explore the expected purity of post-merger samples identified by the CNN. Despite the CNN's good performance in training, the intrinsic rarity of post-mergers leads to a sample that is only ~6 percent pure when the default decision threshold is used. We investigate trade-offs in purity and completeness with a variable decision threshold and find that we recover the statistical distribution of merger-induced star formation rate enhancements. Finally, the performance of the CNN is compared with both traditional automated methods and human classifiers. The CNN is shown to outperform Gini-M20 and asymmetry methods by an order of magnitude in post-merger sample purity on the mock survey data. Although the CNN outperforms the human classifiers on sample completeness, the purity of the post-merger sample identified by humans is frequently higher, indicating that a hybrid approach to classifications may be an effective solution to merger classifications in large surveys.

astro-ph.GA↗

arXiv2021

An astronomical image content-based recommendation system using combined deep learning models in a fully unsupervised mode

We have developed a method that maps large astronomical images onto a two-dimensional map and clusters them. A combination of various state-of-the-art machine learning (ML) algorithms is used to develop a fully unsupervised image quality assessment and clustering system. Our pipeline consists of a data pre-processing step where individual image objects are identified in a large astronomical image and converted to smaller pixel images. This data is then fed to a deep convolutional autoencoder jointly trained with a self-organizing map (SOM). This part can be used as a recommendation system. The resulting output is eventually mapped onto a two-dimensional grid using a second, deep, SOM. We use data taken from ground-based telescopes and, as a case study, compare the system's ability and performance with the results obtained by supervised methods presented by Teimoorinia et al. (2020). The availability of target labels in this data allowed a comprehensive performance comparison between our unsupervised and supervised methods. In addition to image-quality assessments performed in this project, our method can have various other applications. For example, it can help experts label images in a considerably shorter time with minimum human intervention. It can also be used as a content-based recommendation system capable of filtering images based on the desired content.

astro-ph.IM↗

arXiv2021

A re-assessment of strong line metallicity conversions in the machine learning era

Strong line metallicity calibrations are widely used to determine the gas phase metallicities of individual HII regions and entire galaxies. Over a decade ago, based on the Sloan Digital Sky Survey Data Release 4 (SDSS DR4), Kewley \& Ellison published the coefficients of third-order polynomials that can be used to convert between different strong line metallicity calibrations for global galaxy spectra. Here, we update the work of Kewley \& Ellison in three ways. First, by using a newer data release (DR7), we approximately double the number of galaxies used in polynomial fits, providing statistically improved polynomial coefficients. Second, we include in the calibration suite five additional metallicity diagnostics that have been proposed in the last decade and were not included by Kewley \& Ellison. Finally, we develop a new machine learning approach for converting between metallicity calibrations. The random forest algorithm is non-parametric and therefore more flexible than polynomial conversions, due to its ability to capture non-linear behaviour in the data. The random forest method yields the same accuracy as the (updated) polynomial conversions, but has the significant advantage that a single model can be applied over a wide range of metallicities, without the need to distinguish upper and lower branches in $R_{23}$ calibrations. The trained random forest is made publicly available for use in the community.

astro-ph.GA↗

arXiv2020

How do central and satellite galaxies quench? -- Insights from spatially resolved spectroscopy in the MaNGA survey

We investigate how star formation quenching proceeds within central and satellite galaxies using spatially resolved spectroscopy from the SDSS-IV MaNGA DR15. We adopt a complete sample of star formation rate surface densities ($Σ_{\rm SFR}$), derived in Bluck et al. (2020), to compute the distance at which each spaxel resides from the resolved star forming main sequence ($Σ_{\rm SFR} - Σ_*$ relation): $ΔΣ_{\rm SFR}$. We study galaxy radial profiles in $ΔΣ_{\rm SFR}$, and luminosity weighted stellar age (${\rm Age_L}$), split by a variety of intrinsic and environmental parameters. Via several statistical analyses, we establish that the quenching of central galaxies is governed by intrinsic parameters, with central velocity dispersion ($σ_c$) being the most important single parameter. High mass satellites quench in a very similar manner to centrals. Conversely, low mass satellite quenching is governed primarily by environmental parameters, with local galaxy over-density ($δ_5$) being the most important single parameter. Utilising the empirical $M_{BH}$ - $σ_c$ relation, we estimate that quenching via AGN feedback must occur at $M_{BH} \geq 10^{6.5-7.5} M_{\odot}$, and is marked by steeply rising $ΔΣ_{\rm SFR}$ radial profiles in the green valley, indicating `inside-out' quenching. On the other hand, environmental quenching occurs at over-densities of 10 - 30 times the average galaxy density at z$\sim$0.1, and is marked by steeply declining $ΔΣ_{\rm SFR}$ profiles, indicating `outside-in' quenching. Finally, through an analysis of stellar metallicities, we conclude that both intrinsic and environmental quenching must incorporate significant starvation of gas supply.

astro-ph.GA↗

arXiv2020

Assessment of astronomical images using combined machine learning models

We present a two-component Machine Learning (ML) based approach for classifying astronomical images by data-quality via an examination of sources detected in the images and image pixel values from representative sources within those images. The first component, which uses a clustering algorithm, creates a proper and small fraction of the image pixels to determine the quality of the observation. The representative images (and associated tables) are ~800 times smaller than the original images, significantly reducing the time required to train our algorithm. The useful information in the images is preserved, permitting them to be classified in different categories, but the required storage is reduced. The second component, which is a deep neural network model, classifies the representative images. Using ground-based telescope imaging data, we demonstrate that the method can be used to separate 'usable' images from those that present some problems for scientific projects -- such as images that were taken in sub-optimal conditions. This method uses two different data sets as input to a deep model and provides better performance than if we only used the images' pixel information. The method may be used in cases where large and complex data sets should be examined using deep models. Our automated classification approach achieves 97% agreement when compared to classification generated via manual image inspection. We compare our method with traditional results and show that the method improves the results by about 10%, and also presents more comprehensive outcomes.

astro-ph.IM↗

arXiv2020

Comparison of Multi-Class and Binary Classification Machine Learning Models in Identifying Strong Gravitational Lenses

Typically, binary classification lens-finding schemes are used to discriminate between lens candidates and non-lenses. However, these models often suffer from substantial false-positive classifications. Such false positives frequently occur due to images containing objects such as crowded sources, galaxies with arms, and also images with a central source and smaller surrounding sources. Therefore, a model might confuse the stated circumstances with an Einstein ring. It has been proposed that by allowing such commonly misclassified image types to constitute their own classes, machine learning models will more easily be able to learn the difference between images that contain real lenses, and images that contain lens imposters. Using Hubble Space Telescope (HST) images, in the F814W filter, we compare the usage of binary and multi-class classification models applied to the lens finding task. From our findings, we conclude there is not a significant benefit to using the multi-class model over a binary model. We will also present the results of a simple lens search using a multi-class machine learning model, and potential new lens candidates.

astro-ph.GA↗

arXiv2019

The ALMaQUEST Survey: III. Scatter in the resolved star forming main sequence is primarily due to variations in star formation efficiency

Using a sample of 11,478 spaxels in 34 galaxies with molecular gas, star formation and stellar maps taken from the ALMA-MaNGA QUEnching and STar formation (ALMaQUEST) survey, we investigate the parameters that correlate with variations in star formation rates on kpc scales. We use a combination of correlation statistics and an artificial neural network to quantify the parameters that drive both the absolute star formation rate surface density (Sigma_SFR), as well as its scatter around the resolved star forming main sequence (Delta Sigma_SFR). We find that Sigma_SFR is primarily regulated by molecular gas surface density (Sigma_H2) with a secondary dependence on stellar mass surface density (Sigma_*), as expected from an `extended Kennicutt-Schmidt relation'. However, Delta Sigma_SFR is driven primarily by changes in star formation efficiency (SFE), with variations in gas fraction playing a secondary role. Taken together, our results demonstrate that whilst the absolute rate of star formation is primarily set by the amount of molecular gas, the variation of star formation rate above and below the resolved star forming main sequence (on kpc scales) is primarily due to changes in SFE.

astro-ph.GA↗

arXiv2019

Are galactic star formation and quenching governed by local, global or environmental phenomena?

We present an analysis of star formation and quenching in the SDSS-IV MaNGA-DR15, utilising over 5 million spaxels from $\sim$3500 local galaxies. We estimate star formation rate surface densities ($Σ_{\rm SFR}$) via dust corrected $Hα$ flux where possible, and via an empirical relationship between specific star formation rate (sSFR) and the strength of the 4000 Angstrom break (D4000) in all other cases. We train a multi-layered artificial neural network (ANN) and a random forest (RF) to classify spaxels into `star forming' and `quenched' categories given various individual (and groups of) parameters. We find that global parameters (pertaining to the galaxy as a whole) perform collectively the best at predicting when spaxels will be quenched, and are substantially superior to local/ spatially resolved and environmental parameters. Central velocity dispersion is the best single parameter for predicting quenching in central galaxies. We interpret this observational fact as a probable consequence of the total integrated energy from AGN feedback being traced by the mass of the black hole, which is well known to correlate strongly with central velocity dispersion. Additionally, we train both an ANN and RF to estimate $Σ_{\rm SFR}$ values directly via regression in star forming regions. Local/ spatially resolved parameters are collectively the most predictive at estimating $Σ_{\rm SFR}$ in these analyses, with stellar mass surface density at the spaxel location ($Σ_*$) being by far the best single parameter. Thus, quenching is fundamentally a global process but star formation is governed locally by processes within each spaxel.

astro-ph.GA↗

arXiv2019

Deep learning predictions of galaxy merger stage and the importance of observational realism

Machine learning is becoming a popular tool to quantify galaxy morphologies and identify mergers. However, this technique relies on using an appropriate set of training data to be successful. By combining hydrodynamical simulations, synthetic observations and convolutional neural networks (CNNs), we quantitatively assess how realistic simulated galaxy images must be in order to reliably classify mergers. Specifically, we compare the performance of CNNs trained with two types of galaxy images, stellar maps and dust-inclusive radiatively transferred images, each with three levels of observational realism: (1) no observational effects (idealized images), (2) realistic sky and point spread function (semi-realistic images), (3) insertion into a real sky image (fully realistic images). We find that networks trained on either idealized or semi-real images have poor performance when applied to survey-realistic images. In contrast, networks trained on fully realistic images achieve 87.1% classification performance. Importantly, the level of realism in the training images is much more important than whether the images included radiative transfer, or simply used the stellar maps (87.1% compared to 79.6% accuracy, respectively). Therefore, one can avoid the large computational and storage cost of running radiative transfer with a relatively modest compromise in classification performance. Making photometry-based networks insensitive to colour incurs a very mild penalty to performance with survey-realistic data (86.0% with r-only compared to 87.1% with gri). This result demonstrates that while colour can be exploited by colour-sensitive networks, it is not necessary to achieve high accuracy and so can be avoided if desired. We provide the public release of our statistical observational realism suite, RealSim, as a companion to this paper.

astro-ph.GA↗

arXiv2019

Dwarfs or giants? Stellar metallicities and distances in the Canada-France-Imaging-Survey from $ugrizG$ multi-band photometry

We present a new fully data-driven algorithm that uses photometric data from the Canada-France-Imaging-Survey (CFIS; $u$), Pan-STARRS 1 (PS1; $griz$), and Gaia ($G$) to discriminate between dwarf and giant stars and to estimate their distances and metallicities. The algorithm is trained and tested using the SDSS/SEGUE spectroscopic dataset and Gaia photometric/astrometric dataset. At [Fe/H]$<-1.2$, the algorithm succeeds in identifying more than 70% of the giants in the training/test set, with a dwarf contamination fraction below 30% (with respect to the SDSS/SEGUE dataset). The photometric metallicity estimates have uncertainties better than 0.2 dex when compared with the spectroscopic measurements. The distances estimated by the algorithm are valid out to a distance of at least $\sim 80$ kpc without requiring any prior on the stellar distribution, and have fully independent uncertainities that take into account both random and systematic errors. These advances allow us to estimate these stellar parameters for approximately 12 million stars in the photometric dataset. This will enable studies involving the chemical mapping of the distant outer disc and the stellar halo, including their kinematics using the Gaia proper motions. This type of algorithm can be applied in the Southern hemisphere to the first release of LSST data, thus providing an almost complete view of the external components of our Galaxy out to at least $\sim 80$ kpc. Critical to the success of these efforts will be ensuring well-defined spectroscopic training sets that sample a broad range of stellar parameters with minimal biases. A catalogue containing the training/test set and all relevant parameters within the public footprint of CFIS is available online.

astro-ph.GA↗

arXiv2019

CLOVER: Convnet Line-fitting Of Velocities in Emission-line Regions

When multiple star-forming gas structures overlap along the line-of-sight and emit optically thin emission at significantly different radial velocities, the emission can become non-Gaussian and often exhibits two distinct peaks. Traditional line-fitting techniques can fail to account adequately for these double-peaked profiles, providing inaccurate cloud kinematics measurements. We present a new method called Convnet Line-fitting Of Velocities in Emission-line Regions (CLOVER) for distinguishing between one-component, two-component, and noise-only emission lines using 1D convolutional neural networks trained with synthetic spectral cubes. CLOVER utilizes spatial information in spectral cubes by predicting on $3\times3$ pixel sub-cubes, using both the central pixel's spectrum and the average spectrum over the $3\times3$ grid as input. On an unseen set of 10,000 synthetic spectral cubes in each predicted class, CLOVER has classification accuracies of $\sim99\%$ for the one-component class and $\sim97\%$ for the two-component class. For the noise-only class, which is analogous to a signal-to-noise cutoff of four for traditional line-fitting methods, CLOVER has classification accuracy of $100\%$. CLOVER also has exceptional performance on real observations, correctly distinguishing between the three classes across a variety of star-forming regions. In addition, CLOVER quickly and accurately extracts kinematics directly from spectra identified as two-component class members. Moreover, we show that CLOVER is easily scalable to emission lines with hyperfine splitting, making it an attractive tool in the new era of large-scale NH$_3$ and N$_2$H$^+$ mapping surveys.

astro-ph.SR↗

arXiv2019

What shapes a galaxy? - Unraveling the role of mass, environment and star formation in forming galactic structure

We investigate the dependence of galaxy structure on a variety of galactic and environmental parameters for ~500,000 galaxies at z<0.2, taken from the Sloan Digital Sky Survey data release 7 (SDSS-DR7). We utilise bulge-to-total stellar mass ratio, (B/T)_*, as the primary indicator of galactic structure, which circumvents issues of morphological dependence on waveband. We rank galaxy and environmental parameters in terms of how predictive they are of galaxy structure, using an artificial neural network approach. We find that distance from the star forming main sequence (Delta_SFR), followed by stellar mass (M_*), are the most closely connected parameters to (B/T)_*, and are significantly more predictive of galaxy structure than global star formation rate (SFR), or any environmental metric considered (for both central and satellite galaxies). Additionally, we make a detailed comparison to the Illustris hydrodynamical simulation and the LGalaxies semi-analytic model. In both simulations, we find a significant lack of bulge-dominated galaxies at a fixed stellar mass, compared to the SDSS. This result highlights a potentially serious problem in contemporary models of galaxy evolution.

astro-ph.GA↗