Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Photometry of Saturated Stars with Neural Networks

Dominik Winecki, Christopher S. Kochanek

TL;DR

This work tackles the challenge of photometering saturated stars in ASAS-SN by training a multilevel perceptron (MLP) on postage-stamp images to learn detector-specific saturation behavior. The model, trained on ~332k stamps spanning g ≈ 4–14.5 mag and augmented data, achieves near-unbiased photometry with a median offset around -0.004 mag and substantially lower light-curve scatter (≈0.037 mag for saturated stars) than the standard SP2 pipeline. The approach generalizes across cameras and filters via Gaussian-process inter-calibration, delivering smoother light curves for diverse variables (Mira, Cepheid, eclipsing binaries) and offering a faster alternative within Sky Patrol v1.0. Remaining issues stem from ASAS-SN pipeline saturated-star corrections, motivating future work on multi-camera training, using raw images without corrections, and improved training magnitudes; the method promises broader and more reliable bright-star photometry in time-domain surveys.

Abstract

We use a multilevel perceptron (MLP) neural network to obtain photometry of saturated stars in the All-Sky Automated Survey for Supernovae (ASAS-SN). The MLP can obtain fairly unbiased photometry for stars from g~4 to 14~mag, particularly compared to the dispersion (15%-85% 1sigma range around the median) of 0.12 mag for saturated (g<11.5 mag) stars. More importantly, the light curve of a non-variable saturated star has a median dispersion of only 0.037 mag. The MLP light curves are, in many cases, spectacularly better than those provided by the standard ASAS-SN pipelines. While the network was trained on g band data from only one of ASAS-SN's 20 cameras, initial experiments suggest that it can be used for any camera and the older ASAS-SN V band data as well. The dominant problems seem to be associated with correctable issues in the ASAS-SN data reduction pipeline for saturated stars more than the MLP itself. The method is publicly available as a light curve option on ASAS-SN Sky Patrol v1.0.

Photometry of Saturated Stars with Neural Networks

TL;DR

This work tackles the challenge of photometering saturated stars in ASAS-SN by training a multilevel perceptron (MLP) on postage-stamp images to learn detector-specific saturation behavior. The model, trained on ~332k stamps spanning g ≈ 4–14.5 mag and augmented data, achieves near-unbiased photometry with a median offset around -0.004 mag and substantially lower light-curve scatter (≈0.037 mag for saturated stars) than the standard SP2 pipeline. The approach generalizes across cameras and filters via Gaussian-process inter-calibration, delivering smoother light curves for diverse variables (Mira, Cepheid, eclipsing binaries) and offering a faster alternative within Sky Patrol v1.0. Remaining issues stem from ASAS-SN pipeline saturated-star corrections, motivating future work on multi-camera training, using raw images without corrections, and improved training magnitudes; the method promises broader and more reliable bright-star photometry in time-domain surveys.

Abstract

We use a multilevel perceptron (MLP) neural network to obtain photometry of saturated stars in the All-Sky Automated Survey for Supernovae (ASAS-SN). The MLP can obtain fairly unbiased photometry for stars from g~4 to 14~mag, particularly compared to the dispersion (15%-85% 1sigma range around the median) of 0.12 mag for saturated (g<11.5 mag) stars. More importantly, the light curve of a non-variable saturated star has a median dispersion of only 0.037 mag. The MLP light curves are, in many cases, spectacularly better than those provided by the standard ASAS-SN pipelines. While the network was trained on g band data from only one of ASAS-SN's 20 cameras, initial experiments suggest that it can be used for any camera and the older ASAS-SN V band data as well. The dominant problems seem to be associated with correctable issues in the ASAS-SN data reduction pipeline for saturated stars more than the MLP itself. The method is publicly available as a light curve option on ASAS-SN Sky Patrol v1.0.
Paper Structure (5 sections, 2 equations, 11 figures)

This paper contains 5 sections, 2 equations, 11 figures.

Table of Contents

  1. Introduction
  2. Training Data
  3. Methods
  4. Results
  5. Conclusions and Future Work

Figures (11)

  • Figure 1: A log scale image of the postage stamp of a 99.9th percentile bright star with a red contour drawn to show which pixels are used in our model.
  • Figure 2: Final Model Architecture
  • Figure 3: The differences between the input ($g_{in}$) and output ($g_{out}$) magnitudes for the verification stars. The red curves show the median (solid), 68% (dashed) and 95% (dotted) ranges of the differences in bins of $0.25$ mag. A horizontal cyan line is included where the difference is zero.
  • Figure 4: The light curve dispersions of approximately $10^3$ non-variable sources as a function of $g$ magnitude analyzed using the current SP2 pipeline (black triangles) or by the neural network (red squares). The dispersion is defined as one-half of the 15-85% ($1\sigma$) range of the residuals about the median. The saturated (unsaturated) magnitude range is to the left (right) of the vertical line.
  • Figure 5: Light curves of the stars in Fig. \ref{['fig:both']} closest to $g=8$ mag to $11.5$ mag in steps of $0.5$ mag. They were also required to have 68% distribution widths less than $0.053$ mag -- 85% of $g<11.5$ mag stars have smaller dispersions. The horizontal lines are the Gaia-estimated g magnitudes. The number gives the dispersion estimated from the 15-85% ($1\sigma$) range of the points about their median.
  • ...and 6 more figures