Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DRAGNs in the Forest: Identifying Artifacts with Random Forest Models in the VLASS DRAGNs Catalog

Verene Einwalter, Eric J. Hooper, Melissa E. Morris, Sarah Bach, Yjan A. Gordon

TL;DR

This paper tackles artifact contamination in the VLASS DRAGNs catalog by training random forest classifiers to count the number of artifacts in DRAGNhunter-identified multi-component sources. It leverages features derived from DRAGNhunter outputs, applies careful training-set selection (notably a log-log LAS S/N and flux S/N sampling approach), and demonstrates that a doubles classifier built on triples-trained data achieves high accuracy. The best-performing model, a log-log doubles classifier, yields a weighted F1 score of $97.0\%$ on validation, with bootstrap estimates giving $97.01\%^{+1.12\%}_{-1.32\%}$, and reports completeness of $99.3\%$ and purity of $97.7\%$ for an artifact-free subset. The approach provides probabilistic artifact classifications, enables a more complete and purer DRAGN catalog than the traditional DRAGNhunter filter, and offers a scalable framework for artifact identification in current and future radio surveys.

Abstract

The Quick Look data products from the Very Large Array Sky Survey (VLASS) contain widespread imaging artifacts arising from the simplified imaging algorithm used in their production. The catalog of double radio sources associated with active galactic nuclei (DRAGNs) found in the VLASS first epoch Quick Look release using the DRAGNhunter algorithm suffers from contamination from these artifacts. These sources contain two or three individual components, each of which can be an artifact. We train random forest models to classify these DRAGNs based on the number of artifacts they contain, ranging from zero to three artifacts. We optimize our models and mitigate the class imbalance of our dataset with judicious training set selection, and the best of our models achieves a weighted F1 score of $97.01\%^{+1.12\%}_{-1.32\%}$. Using our classifications, we produce a catalog of VLASS DRAGNs from which an estimated 99.3% complete catalog of 97.7% artifact-free sources can be extracted.

DRAGNs in the Forest: Identifying Artifacts with Random Forest Models in the VLASS DRAGNs Catalog

TL;DR

This paper tackles artifact contamination in the VLASS DRAGNs catalog by training random forest classifiers to count the number of artifacts in DRAGNhunter-identified multi-component sources. It leverages features derived from DRAGNhunter outputs, applies careful training-set selection (notably a log-log LAS S/N and flux S/N sampling approach), and demonstrates that a doubles classifier built on triples-trained data achieves high accuracy. The best-performing model, a log-log doubles classifier, yields a weighted F1 score of on validation, with bootstrap estimates giving , and reports completeness of and purity of for an artifact-free subset. The approach provides probabilistic artifact classifications, enables a more complete and purer DRAGN catalog than the traditional DRAGNhunter filter, and offers a scalable framework for artifact identification in current and future radio surveys.

Abstract

The Quick Look data products from the Very Large Array Sky Survey (VLASS) contain widespread imaging artifacts arising from the simplified imaging algorithm used in their production. The catalog of double radio sources associated with active galactic nuclei (DRAGNs) found in the VLASS first epoch Quick Look release using the DRAGNhunter algorithm suffers from contamination from these artifacts. These sources contain two or three individual components, each of which can be an artifact. We train random forest models to classify these DRAGNs based on the number of artifacts they contain, ranging from zero to three artifacts. We optimize our models and mitigate the class imbalance of our dataset with judicious training set selection, and the best of our models achieves a weighted F1 score of . Using our classifications, we produce a catalog of VLASS DRAGNs from which an estimated 99.3% complete catalog of 97.7% artifact-free sources can be extracted.
Paper Structure (22 sections, 3 equations, 14 figures)

This paper contains 22 sections, 3 equations, 14 figures.

Table of Contents

  1. Introduction
  2. Motivation
  3. Artifact Visual Classification
  4. Preliminary Clustering
  5. Analogy Between Triple and Double Artifact Classes
  6. Random Forest Methodology
  7. Data and Model Preparation
  8. Determining Final Artifact Classification Classes
  9. 1-Artifact Triple Sources
  10. Training Set Selection
  11. Random Forest Classification of Doubles
  12. Classification Models
  13. Triples-Trained Model
  14. Log-Log LAS S/N and Flux S/N Selected Model
  15. Doubles Classification Performance and Model Comparison
  16. ...and 7 more sections

Figures (14)

  • Figure 1: Collage of 1.5'x1.5' VLASS images of triple-component DRAGNs identified by DRAGNhunter. The ellipses denote components as identified by DRAGNhunter, where the green ellipses denote the lobe or jet hot spot component, the cyan ellipse denotes the identified core, and the green X denotes the AllWISE host as identified in gordon_quick_2023, if one was found. (Top line) Examples of what typical DRAGN triples with 3 artifacts look like. Note that they tend to result from artifacts around very bright point sources (e.g. a, b), which are usually single sources, but can also be very bright jet lobes from extended sources (e.g. c). (Center Line) Examples of what typical 2-artifact triples look like. These are predominantly characterized as 2 artifacts surrounding an unresolved point source. (Bottom Line) Example 1-artifact triple sources. These are usually double sources with a spurious artifact component around a bright jet lobe.
  • Figure 2: Examples of triple sources with dubious morphologies which are difficult to classify. The ellipses denote components as identified by DRAGNhunter, where the green ellipses denote the lobe or jet hot spot component, the cyan ellipse denotes the identified core, and the green X denotes the AllWISE host as identified in gordon_quick_2023, if one was found. (a) A double with a strong Y-shaped sidelobe pattern where the core and bottom component are artifacts. (b) A source where part of one highly elongated component intersects, but does not encompass real emission. Components like this, which are centered on an artifact and happen to overlap with real emission, were classified as artifacts. (c) A source with prominent, bright sidelobes. (d) The diffuse, extended emission of this AGN is faint and difficult to distinguish from background noise, however all components are associated with real, separate emission.
  • Figure 3: Confusion matrix comparing the results of visual inspection for artifacts of this paper and spurious detections of the triples in gordon_quick_2023, where 1 denotes spurious and 0 denotes not spurious. Spurious sources are those that contain artifacts. The percentages and the color of each square are determined by the fraction of the total population of the row, i.e. the fraction of the true class, that is present in each square. All sources that we identified in this paper as having 1 or more artifacts were labeled as 1 in this matrix.
  • Figure 4: Histogram of the prominence of the brightest component of the triples identified in the DRAGNhunter catalog. 1-artifact sources are not included because the shape of their histogram is flat and cannot be cleanly separated from those of the other artifact classes.
  • Figure 5: (a) Scatterplot of LAS S/N vs. Flux S/N for all triples grouped by number of artifacts in each source as identified by visual inspection. This particular set of parameters shows that the 0-, 2-, and 3-artifact classes of sources cluster in approximately 3 separate areas. (b) Scatterplot of LAS S/N vs. Flux S/N for all doubles which suggests that the double sources may also cluster into artifact classes within this parameter space like the triples.
  • ...and 9 more figures