Type: Poster
Session: Posters (Monday & Tuesday)
Author: Andreia Dourado

Abstract: Obtaining robust photometric redshift estimates for Rubin LSST extragalactic objects is fundamental for several key objectives across scientific areas. While official redshift products will be available to the public and the Science Collaborations, the astronomical community should be able to generate, validate, and distribute photometric redshift catalogs suitable for cutting-edge science with Rubin data sets. In this project, we perform an end-to-end production run of a photometric redshift catalog on Rubin DP0.2 data. With data obtained from the Rubin Science Platform, we employ the infrastructure of Brazil's in-kind contribution to perform training, validation, and production of photometric redshifts with TPz, a well-known astronomical implementation of random forest regression. We use the RAIL infrastructure on the LIneA Apollo cluster to complete these tasks. In the process, we analyze individual and ensemble metrics to assess the performance of the code, identify potential issues, and optimize the choices made for some of the use cases in cosmology.