New results in applying the machine learning to GRB redshift estimation

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

“…Another problem, hardly explored in the literature (besides our study), is to account for errors of GRB variables used to train ML models. Further, comparing our results with Rácz et al (2017), who achieved a r = 0.67, we see a 22% increase in our correlation in the ( ) + z log 1 10 scale when we apply the bias correction. In addition, our methodology is more complete than this work, since we use the LASSO feature selection, the M-estimator, the nested 100 10fCV, the SuperLearner, and the bias correction.…”

“…In comparison to other attempts to infer GRB redshift, our ensemble achieves an increase of 63% and 38% in the correlation between predicted and observed redshift reaching r = 0.93, compared to other works in which only random forest or gradient boosting alone was used. Ukwatta et al (2016) found r = 0.57 with random forest, while Rácz et al (2017) found r = 0.67 both with random forest and gradient boosting. The main difference, besides our enhanced prediction and the fact that we use a more complete methodology, which has been detailed in Section 3, is the use of the plateau properties.…”

New detailed characterization of the residual luminescence emitted by the GAGG:Ce scintillator crystals for the HERMES Pathfinder mission

“…The advent of third-generation gravitational wave detectors and modern astronomical facilities anticipates the occurrence of numerous multimessenger events with similar characteristics. In [68], the authors employ In [209], the authors investigated GRBs and their afterglows using three instruments on board Swift, operating in gamma-ray (BAT), X-ray (XRT), ultraviolet, and optical (UVOT) wavebands. Instead of solely relying on the Swift GRB Table from the NASA website (as in [251]), they combined this table with the Swift-XRT GRB Catalogue from the UK Swift Science Data Centre to create a new table that includes all X-ray spectral fitting data, offering higher precision compared to the data in the Swift GRB Table . With these enriched features, they applied both XGBoost and Random Forest regression models to estimate the redshift of the GRBs.…”