Predicting the Redshift of Gamma-Ray Loud AGNs Using Supervised Machine Learning. II

Abstract

Measuring the redshift of active galactic nuclei (AGNs) requires the use of time-consuming and expensive spectroscopic analysis. However, obtaining redshift measurements of AGNs is crucial as it can enable AGN population studies, provide insight into the star formation rate, the luminosity function, and the density rate evolution. Hence, there is a requirement for alternative redshift measurement techniques. In this project, we aim to use the Fermi Gamma-ray Space Telescope’s 4LAC Data Release 2 catalog to train a machine-learning (ML) model capable of predicting the redshift reliably. In addition, this project aims at improving and extending with the new 4LAC Catalog the predictive capabilities of the ML methodology published in Dainotti et al. Furthermore, we implement feature engineering to expand the parameter space and a bias correction technique to our final results. This study uses additional ML techniques inside the ensemble method, the SuperLearner, previously used in Dainotti et al. Additionally, we also test a novel ML model called Sorted L-One Penalized Estimation. Using these methods, we provide a catalog of estimated redshift values for those AGNs that do not have a spectroscopic redshift measurement. These estimates can serve as a redshift reference for the community to verify as updated Fermi catalogs are released with more redshift measurements.

Classification of Fermi-LAT unidentified gamma-ray sources using catboost gradient boosting decision trees

“…When comparing the predictions on the test set, we obtained a R2 score of 0.56 with a 0.46 RMSE. The predicted and real redshifts presented a Pearson correlation of 0.71, similar to previous works Dainotti et al (2021); Narendra et al (2022).…”

“…In Figure 1, we plot a scatter relation between the predicted and real redshifts for the 4LAC test set. The Pearson correlation coefficient between prediction and actual redshifts is 𝑟 = 0.71, which, despite the different approach and newer data release, is a very similar value to those found by Dainotti et al (2021) and Narendra et al (2022). The average redshift value is found to be 𝑧 𝑎𝑣𝑔 =0.63, with a maximum 𝑧 𝑚𝑎𝑥 =1.97.…”

“…Previous works have aimed to predict the redshift of gammaray AGNs with machine learning techniques, such as a Superlearner combination of XGBoost, RF, Bayes GLM, and Big LASSO in 4LAC Dainotti et al (2021), Hierarchical Correlation Reconstruction also in 4LAC Duda (2022) or SLOPE in 4LAC-DR2 Narendra et al (2022), with a Pearson correlation coefficient of 0.74 and a RMSE of 0.467.…”

Progenitors of Low-redshift Gamma-Ray Bursts

“…We note that the sample of SGRBs used inDainotti et al (2021d) also includes the short with extended emission (SEE). The total sample comprises 68 GRBs, with 41 SGRBs and 27 SEEs assuming that SEEs have the same progenitors as the SGRBs following the discussion ofBarkov & Pozanenko (2011).13 Future analysis with machine learning will allow the estimate of larger sample of GRBs; seeDainotti et al (2021b),Narendra et al (2022).14 This sample stems from a large compilation byWang et al (2020) spanning the period from GRB 910421 to GRB 160509A.…”

Inferring the Redshift of More than 150 GRBs with a Machine-learning Ensemble Model

“…This results in a Gaussian distribution with a mean = 0.48 and a standard deviation = 0.128 for this response variable, rather than a distribution with tails as shown for the distribution of the redshift in the scatter matrix plot of Figure 2. This new response variable is chosen similarly as in previous literature Narendra et al 2022), and it is a natural choice since it mimics the evolution of the variables. In addition, z + 1 is a more natural parameterization of the cosmological variable z.…”