Autonomous Gaussian Decomposition

Lindner, Robert R.; Vera-Ciro, Carlos; Murray, Claire E.; Stanimirović, Snežana; Babler, B.; Heiles, Carl; Hennebelle, P.; Goss, W. M.; Dickey, John M.

doi:10.1088/0004-6256/149/4/138

Cited by 79 publications

(102 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Armed with synthetic 21 cm emission and absorption profile data created from the 3D hydrodynamical simulations from KOK14 and high-sensitivity H I observations from 21-SPONGE, we address two main questions: (1) how well do H I spectral lines and our analysis methods recover simulated properties of interstellar gas structures; and (2) how do simulated H I spectra compare with real observations? To analyze 9355 synthetic and 52 real observations in an unbiased and uniform way, we apply the AGD algorithm (Lindner et al 2015) identically to both data sets. With these fits in hand, we compare simulated properties of gas structures along each LOS with observed properties of the Gaussian components.…”

Section: Discussionmentioning

confidence: 99%

“…We began by constructing a synthetic H I data set from the Gaussian components detected by the Millennium Arecibo 21 cm Absorption Line Survey (HT03, Heiles & Troland 2003b). The synthetic training data set construction and training are described fully in Lindner et al (2015), and summarized here for clarity. We selected the number of components in each synthetic spectrum to be a uniform random integer ranging from the mean number of components in the survey (3) to the maximum number (8; HT03), and then drew the component parameters from the published HT03 amplitude, full width at half maximum (FWHM), and mean velocity distributions with replacement.…”

Section: Gaussian Decomposition With Agdmentioning

confidence: 99%

“…To analyze the synthetic spectral profiles, we use the Autonomous Gaussian Decomposition (AGD) algorithm (Lindner et al 2015). AGD implements derivative-based computer vision to perform Gaussian decomposition of spectral lines, enabling efficient, reproducible, and objective spectral decomposition.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Recovering Interstellar Gas Properties with Hi Spectral Lines: A Comparison between Synthetic Spectra and 21-SPONGE

Murray

Stanimirović

Kim

et al. 2017

ApJ

Self Cite

View full text Add to dashboard Cite

We analyze synthetic neutral hydrogen (H I) absorption and emission spectral lines from a high-resolution, threedimensional hydrodynamical simulation to quantify how well observational methods recover the physical properties of interstellar gas. We present a new method for uniformly decomposing H I spectral lines and estimating the properties of associated gas using the Autonomous Gaussian Decomposition (AGD) algorithm. We find that H I spectral lines recover physical structures in the simulation with excellent completeness at high Galactic latitude, and this completeness declines with decreasing latitude due to strong velocity-blending of spectral lines. The temperature and column density inferred from our decomposition and radiative transfer method agree with the simulated values within a factor of <2 for the majority of gas structures. We next compare synthetic spectra with observations from the 21-SPONGE survey at the Karl G. Jansky Very Large Array using AGD. We find more components per line of sight in 21-SPONGE than in synthetic spectra, which reflects insufficient simulated gas scale heights and the limitations of local box simulations. In addition, we find a significant population of lowoptical depth, broad absorption components in the synthetic data which are not seen in 21-SPONGE. This population is not obvious in integrated or per-channel diagnostics, and reflects the benefit of studying velocityresolved components. The discrepant components correspond to the highest spin temperatures ( < < T 1000 4000 K s ), which are not seen in 21-SPONGE despite sufficient observational sensitivity. We demonstrate that our analysis method is a powerful tool for diagnosing neutral interstellar medium conditions, and future work is needed to improve observational statistics and implementation of simulated physics.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Gaussian Decomposition With Agdmentioning

confidence: 99%

See 1 more Smart Citation

Recovering Interstellar Gas Properties with Hi Spectral Lines: A Comparison between Synthetic Spectra and 21-SPONGE

Murray

Stanimirović

Kim

et al. 2017

ApJ

Self Cite

View full text Add to dashboard Cite

show abstract

“…Lindner et al (2015) developed a method of autonomous Gaussian decomposition for the 21 cm SPectral line Observations of Neutral Gas with the EVLA survey (Murray et al 2015). Their technique was centred on determining the best initial guesses for the Gaussian fitting, which is often the most difficult part of line fitting (see Ho et al 2016b for a discussion on selecting initial conditions for emission line spectra) and used a combination of computer vision (often an ANN method) and derivative spectra.…”

Section: A S U P E Rv I S E D a Rt I F I C I A L N E U R A L N E T W mentioning

confidence: 99%

“…Our study, unlike Lindner et al (2015), centres on what comes after the fitting process. Instead of determining the number of Gaussians beforehand, we make the determination after the emission line fitting has been conducted with 1-, 2-and 3-Gaussian components.…”

Section: A S U P E Rv I S E D a Rt I F I C I A L N E U R A L N E T W mentioning

confidence: 99%

Using an artificial neural network to classify multicomponent emission lines with integral field spectroscopy from SAMI and S7

Hampton

Medling

Groves

et al. 2017

Monthly Notices of the Royal Astronomical Society

View full text Add to dashboard Cite

Integral field spectroscopy (IFS) surveys are changing how we study galaxies and are creating vastly more spectroscopic data available than before. The large number of resulting spectra makes visual inspection of emission line fits an infeasible option. Here, we present a demonstration of an artificial neural network (ANN) that determines the number of Gaussian components needed to describe the complex emission line velocity structures observed in galaxies after being fit with LZIFU. We apply our ANN to IFS data for the S7 survey, conducted using the Wide Field Spectrograph on the ANU 2.3 m Telescope, and the SAMI Galaxy Survey, conducted using the SAMI instrument on the 4 m Anglo-Australian Telescope. We use the spectral fitting code LZIFU (Ho et al. 2016a) to fit the emission line spectra of individual spaxels from S7 and SAMI data cubes with 1-, 2-and 3-Gaussian components. We demonstrate that using an ANN is comparable to astronomers performing the same visual inspection task of determining the best number of Gaussian components to describe the physical processes in galaxies. The advantage of our ANN is that it is capable of processing the spectra for thousands of galaxies in minutes, as compared to the years this task would take individual astronomers to complete by visual inspection.

show abstract

How reliable is distribution of relaxation times (DRT) analysis? A dual regression-classification perspective on DRT estimation, interpretation, and accuracy

Huang

Sullivan

Zakutayev

et al. 2023

Electrochimica Acta

View full text Add to dashboard Cite

Autonomous Gaussian Decomposition

Cited by 79 publications

References 46 publications

Recovering Interstellar Gas Properties with Hi Spectral Lines: A Comparison between Synthetic Spectra and 21-SPONGE

Recovering Interstellar Gas Properties with Hi Spectral Lines: A Comparison between Synthetic Spectra and 21-SPONGE

Using an artificial neural network to classify multicomponent emission lines with integral field spectroscopy from SAMI and S7

How reliable is distribution of relaxation times (DRT) analysis? A dual regression-classification perspective on DRT estimation, interpretation, and accuracy

Contact Info

Product

Resources

About