Star cluster classification in the PHANGS–<i>HST</i> survey: Comparison between human and machine learning approaches

Whitmore, Bradley C.; Lee, Janice; Chandar, Rupali; Thilker, David A.; Hannon, Stephen; Wei, Wei; Huerta, E. A.; Bigiel, Frank; Boquien, M.; Chevance, Mélanie; Dale, Daniel A.; Deger, Sinan; Grasha, Kathryn; Klessen, Ralf S.; Kruijssen, J. M. Diederik; Larson, Kirsten L.; Mok, Angus; Rosolowsky, Erik; Schinnerer, E.; Schruba, Andreas; Úbeda, Leonardo; Dyk, Schuyler D. Van; Watkins, Elizabeth J.; Williams, Thomas G.

doi:10.1093/mnras/stab2087

Cited by 36 publications

(38 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To counter confusion from nearby objects, observational surveys that rely on aperture photometry use as small apertures as possible and aim at building their star cluster catalogues with emphasis on single, symmetric and, for example, uniformly colored cluster candidates (Cook et al 2012;Adamo et al 2017). The classification of bona fide clusters can be either computer or human generated, or a combination of both, at varying level of agreement between the two (Whitmore et al 2021). Such objects may be therefore extremely challenging to recover in a reliable manner in the 2D projections of intensely star forming regions and often the central star forming knots and clumpy cluster candidates are left out of the analysis all together (Larsen 2002;Adamo et al 2020).…”

Section: Introductionmentioning

confidence: 99%

A panchromatic view of star cluster formation in a simulated dwarf galaxy starburst

Lahén,

Naab,

Kauffmann

2021

Preprint

View full text Add to dashboard Cite

We present a photometric analysis of star and star cluster (SC) formation in a high-resolution simulation of a dwarf galaxy starburst that allows the formation of individual stars to be followed. Previous work has demonstrated that the properties of the SCs formed in the simulation are in good agreement with observations. In this paper, we create mock spectral energy distributions and broad-band photometric images using the radiative transfer code 9. We test several observational star formation (SF) tracers and find that 24 𝜇m, total infrared and H 𝛼 trace the underlying SF rate during the (post)starburst phase, while UV tracers yield a more accurate picture of star formation during quiescent phases prior to and after the merger. We then place the simulated galaxy at distances of 10 and 50 Mpc and use aperture photometry at Hubble Space Telescope resolution to analyse the simulated SC population. During the starburst phase, a hierarchically forming set of SCs leads inaccurate source separation because of crowding. This results in estimated SC mass function slopes that are up to ∼ 0.3 shallower than the true slope of ∼ −2 found for the bound clusters identified from the particle data in the simulation. The masses of the largest clusters are overestimated by a factor of up to 2.5 due to unresolved clusters within the apertures. The aperture-based analysis also produces a relation between cluster formation efficiency and SF rate surface density that is flatter than that recovered from bound clusters. The differences are strongest in quiescent SF environments.

show abstract

Section: Introductionmentioning

confidence: 99%

A panchromatic view of star cluster formation in a simulated dwarf galaxy starburst

Lahén,

Naab,

Kauffmann

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…We build training sets of ∼15 hand-selected stars, clusters, and galaxies to help guide the automated cuts. Following Whitmore et al (2021), we identify our training set using a variety of tools, including color images and plotting radial profiles using the imexam python package, where stars all have similar sharp radial profiles but clusters are clearly more extended. Our primary quantitative tool is the Concentration Index (CI), which is the difference in the V-magnitude measured in apertures of 0.5 and 3 pixel radii.…”

Section: Cluster Selectionmentioning

confidence: 99%

The Ancient Globular Clusters of NGC 1291

Hixenbaugh¹,

Chandar²,

Mok³

2022

Self Cite

View full text Add to dashboard Cite

We present a new catalog of 81 ancient globular clusters (GCs) in the early-type spiral (SB0/a) galaxy NGC 1291. Candidates have been selected from B, V, and I band images taken with the Hubble Space Telescope, which also reveal 17 younger (τ ≲ few × 100 Myr) clusters. The luminosity function shows a peaked shape similar to that found for GC systems in other spiral and elliptical galaxies. The ancient clusters have a bimodal color distribution, with approximately 65% (35%) of the population having blue (red) colors. The red, presumably metal-rich, GCs are more centrally concentrated, as expected for a bulge population; while the blue, presumably metal-poor, GCs are more broadly distributed, consistent with expectations of a halo population. The specific frequency of GCs in NGC 1291 is higher than found previously in most spiral galaxies. However, if we consider just the blue subpopulation, we find S N,blue = 0.50 ± 0.06, quite similar to that found for other spirals. This result supports the hypothesis of a universal population of halo GCs in spirals. The fraction of red GCs in NGC 1291, when compared with those found in other galaxies, suggests that these correlate with host galaxy type rather than with host galaxy luminosity.

show abstract

“…They report a TPR of ∼81% and a FDR of ∼19% on a cluster/non-cluster classification task, reaching a TPR of ∼93% and a FDR of ∼7% if testing is restricted to high-mass objects (153 sources). Whitmore et al (2021) report a TPR of ∼82% for a similar cluster/non-cluster classification task for sources from five galaxies of the PHANGS-HST sample using two popular CNN architectures (Resnet18 and VGG19-BN) trained via transfer learning on ∼5500 sources of 10 LEGUS galaxies. When only applying their trained models to isolated objects and for detecting compact and symmetric star clusters, they reach a TPR of ∼92%.…”

Section: Model Performance and Generalisation To New Datamentioning

confidence: 99%

“…For the latter case, the catalogues of young star clusters created within the LEGUS or PHANGS projects (e.g. Whitmore et al 2021;Pérez et al 2021) could be used to build a comprehensive sample of GCs and young star clusters for training models. However, when including young star clusters, possibly additional features that encompass the specifics of young star clusters in comparison to mostly spherical GCs need to be included (see e.g Whitmore et al 2011or Deger et al 2022.…”

Section: Model Performance and Generalisation To New Datamentioning

confidence: 99%

Evaluating the feasibility of interpretable machine learning for globular cluster detection

Dold,

Fahrion

2022

Preprint

View full text Add to dashboard Cite

Extragalactic globular clusters (GCs) are important tracers of galaxy formation and evolution because their properties, luminosity functions, and radial distributions hold valuable information about the assembly history of their host galaxies. Obtaining GC catalogues from photometric data involves several steps which will likely become too time-consuming to perform on the large data volumes that are expected from upcoming wide-field imaging projects such as Euclid. In this work, we explore the feasibility of various machine learning methods to aid the search for GCs in extensive databases. We use archival Hubble Space Telescope data in the F475W and F850LP bands of 141 early-type galaxies in the Fornax and Virgo galaxy clusters. Using existing GC catalogues to label the data, we obtain an extensive data set of 84929 sources containing 18556 GCs and we train several machine learning methods both on image and tabular data containing physically relevant features extracted from the images. We find that our evaluated machine learning models are capable of producing catalogues of similar quality as the existing ones which were constructed from mixture modelling and structural fitting. The best performing methods, ensemble-based models like random forests and convolutional neural networks, recover ∼ 90 − 94 % of GCs while producing an acceptable amount of false detections (∼ 6 − 8%) -with some falsely detected sources being identifiable as GCs that have not been labelled as such in the used catalogues. In the magnitude range 22 < m4_g ≤ 24.5 mag, 98 -99% of GCs are recovered. We even find such high performance levels when training on Virgo and evaluating on Fornax data (and vice versa), illustrating that the models are transferable to environments with different conditions such as different distances than in the used training data. Apart from performance metrics, we demonstrate how interpretable methods can be utilised to better understand model predictions, recovering that magnitudes, colours, and sizes are important properties for identifying GCs. Moreover, comparing colour distributions from our detected sources to the reference distributions from input catalogues finds great agreement and the mean colour is recovered even for systems with less than 20 GCs. These are encouraging results, indicating that similar methods trained on an informative subsample can be applied for creating GC catalogues for a large number of galaxies, with tools being available for increasing the transparency and reliability of said methods.

show abstract

Star cluster classification in the PHANGS–HST survey: Comparison between human and machine learning approaches

Cited by 36 publications

References 46 publications

A panchromatic view of star cluster formation in a simulated dwarf galaxy starburst

A panchromatic view of star cluster formation in a simulated dwarf galaxy starburst

The Ancient Globular Clusters of NGC 1291

Evaluating the feasibility of interpretable machine learning for globular cluster detection

Contact Info

Product

Resources

About