Performance of the network NetZ_LRG (bottom) trained on LRGs only. Left-hand side: histograms of the redshift distributions in red we show the distribution of the reference redshifts used to train the network (ground truth) and in blue the predicted redshift distribution. Right panel: a 1:1 comparison of reference and predicted redshifts. The red lines show the median and the gray bands the 1σ and 2σ confidence levels.