Machine Learning Methods for High-Redshift Quasar Detection with Selection Bias Mitigation.

• Grant number: 25/04853-5
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: September 01, 2025
• End date: August 31, 2026
• Field of knowledge: Physical Sciences and Mathematics - Probability and Statistics - Applied Probability and Statistics
• Principal Investigator: Rafael Izbicki
• Grantee: Bruno Marcondes e Resende
• Host Institution: Centro de Ciências Exatas e de Tecnologia (CCET). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil

Abstract

Redshift is an astronomical measure that indicates the shift of an electromagnetic wave toward the red spectrum, used to determine the distance of astronomical objects such as galaxies and quasars. The greater the redshift, the more distant and older the object is. Detecting high-redshift quasars is crucial for understanding the formation of supermassive black holes in the early universe. Traditionally, redshift is measured through spectroscopy, a time-consuming and costly process. Photometry, which measures the intensity of light in different wavelength bands, combined with machine learning (ML), is used to estimate redshift more quickly. However, there is the challenge of selection bias, which results in training datasets that underrepresent quasars typically detected through spectroscopy.This project aims to develop machine learning methods that mitigate this bias, enabling more efficient detection of high-redshift quasars. Using data from the S-PLUS (Southern Photometric Local Universe Survey) photometric survey, which covers 9,300 square degrees of the southern celestial hemisphere and employs a photometric system with 12 band-pass filters, the project seeks not only to improve the accuracy of redshift estimates but also to identify candidates for very high-redshift quasars. This will deepen the understanding of the early universe and the mechanisms that led to the formation of supermassive black holes.