Notes. Columns are: setup identifier, number of hierarchical clusters, resolution on-sky, and systematic-tracers for training. Various configurations for our SOM and organised randoms-creation pipeline. Throughout the text, we refer to these by their setup ID. This shorthand typically just indicates the number of hierarchical clusters defined on the SOM, e.g., 800 or 100, and the number of systematic-tracer variables used in training: A is just the PSF and threshold parameters, B then includes the r-band magnitude limit, and C further includes Galactic dust extinction and stellar density (all described in Table 1). The exception T1mock refers to the SOM trained against artificial systematic-tracers A1-3, B1-3, and C1-2, which are described in Sect. 4.1. N_HC is the number of hierarchical clusters to be defined on a given SOM, ‘res (smth)’ are the resolution and Gaussian smoothing width in arcminutes of the sky-grid used to group galaxies and populate random fields, and ‘systematics’ refers to the set of systematic-tracer variables presented to each SOM in training (see Table 1). All SOMs employed in our analysis have dimensions 100 × 100.