Hyper Suprime-Cam Structural Parameter Catalog

Morphological Parameters and Associated Uncertainties for 8 Million Galaxies in the Hyper Suprime-Cam Wide Survey

Image showing potential science impacts of our HSC structural parameter catalog Highlights of our HSC structural parameter catalog shown along with science cases we are currently pursuing using this catalog.

Dataset & Algorithm Used

We used the Galaxy Morphology Posterior Estimation Network (GaMPEN) to estimate the structural parameters (bulge-to-total light ratio, effective radius, and total magnitude) and associated uncertainties for all galaxies in the Hyper Suprime-Cam (HSC) Wide survey with $z \leq 0.75$ and $m \leq 23$.

Image showing a schematic diagram of GaMPEN Schematic diagram of GaMPEN. GaMPEN is a machine learning framework that estimates Bayesian posteriors of (user-defined) galaxy structural parameters for arbitrarily large numbers of galaxies.

We use different imaging bands across different redshift ranges to trace the same ground-frame wavelength. The breakdown of our sample into the three different redshift bins is shown below.

Image showing the breakdown of sample across the different redshift slices.

Training GaMPEN using $ < 1 \% $ of our Data

We first train GaMPEN on simulations of galaxies and then fine-tune the already trained network using $ <1 \% $ of our total dataset for training. This is an important demonstration that ML frameworks can be used to measure galaxy properties in new surveys, which do not have already-classified large training sets readily available. Our implemented two-step process provides a new framework that can be easily used for upcoming large imaging surveys like the Rubin-LSST, Euclid, and NGRST.

Image depicting the two stage process we use to train GaMPEN.

Accuracy of GaMPEN's Predictions

We compare the GaMPEN's predictions against ground-truth values in simulations, as well as the predictions of traditional light-profile fitting codes. GaMPEN's measurements are consistent overall and outperform those of GALFIT for galaxies with $R_e \leq 2$" ($\sim 12.6$ kpc at $z=0.5$)

GaMPEN's predicted values against ground truth is show below for a set of simulated HSC galaxies

The true values of the galaxy parameters plotted against the most probable values predicted by \gampen{}.

We show below a few examples of applying GaMPEN on real HSC galaxies

Image showing posterior distributions predicted by GaMPEN for different galaxies.

Accuracy of GaMPEN's Posteriors

By comparing our results to those obtained using light-profile fitting, we demonstrate that GaMPEN's predicted posterior distributions are well-calibrated ($\lesssim 5\%$ deviation) and accurate. This represents a significant improvement over light profile fitting algorithms which underestimate uncertainties by as much as $\sim60\%$.

Below, we plot the percentile coverage probabilities for the $68.27\%$ confidence interval obtained by GaMPEN on our HSC sample compared to coverage probabilities obtained by various light-profile fitting algorithms on simulated Euclid data. The coverage probabilities are defined as the percentage of the total test examples where the value determined using light profile fitting lies within a particular confidence interval of the predicted distribution.

Percentile coverage probabilities for the $68.27\%$ confidence interval obtained by \gampen{} on our HSC sample compared to coverage probabilities obtained by various light-profile fitting algorithms on simulated Euclid data The rightmost set of bars shows the values calculated on the entire dataset, while the other sets display values calculated on sub-samples of galaxies with specific magnitude ranges (AB mag, shown on the x-axis). As this figure shows, Compared to light-profile fitting tools, the uncertainties predicted by GaMPEN are better calibrated by $\sim15-25\%$ overall and by as much as $\sim60\%$ for the brightest galaxies.