Speaker
Description
Galaxy clusters are key tracers of the large-scale structure of the Universe and powerful probes of cosmology. However, identifying clusters directly in redshift space remains challenging because peculiar velocities distort their apparent distribution, producing anisotropic structures such as Finger-of-God elongations and coherent infall. In this talk, I will present preliminary results from a machine-learning approach designed to identify galaxy clusters in simulated redshift-space galaxy catalogues. The method treats galaxies as three-dimensional point-cloud data and aims to learn cluster-like environments directly from their spatial distribution, without requiring a regular grid representation. I will discuss the construction of simulated samples, the labelling strategy, and the performance of the current classification pipeline. This work is intended as a first step toward developing flexible cluster-identification methods for realistic galaxy surveys, where redshift-space distortions, projection effects, and variable cluster richness must be handled directly.